Issues In The Study Of Pidgin And Creole Languages. (Studies in Language Companion)

Issues in the Study of Pidgin and Creole Languages Studies in Language Companion Series (SLCS) The SLCS series has be...

Author: Claire Lefebvre

80 downloads 1019 Views 14MB Size Report

This content was uploaded by our users and we assume good faith they have the permission to share this book. If you own the copyright to this book and it is wrongfully on our website, we offer a simple DMCA procedure to remove your content from our site. Start by pressing the button below!

Report copyright / DMCA form

DOWNLOAD PDF

Issues in the Study of Pidgin and Creole Languages

Studies in Language Companion Series (SLCS) The SLCS series has been established as a companion series to Studies in Language, International Journal, sponsored by the Foundation “Foundations of Language”.

Series Editors Werner Abraham

Michael Noonan

University of Vienna

University of Wisconsin, Milwaukee

Editorial Board Joan Bybee

Ekkehard König

University of New Mexico

Free University of Berlin

Ulrike Claudi

Christian Lehmann

University of Cologne

University of Erfurt

Bernard Comrie

Robert Longacre

Max Planck Institute For Evolutionary Anthropology, Leipzig

University of Texas, Arlington

William Croft

Carnegie-Mellon University

University of Manchester

Marianne Mithun

Östen Dahl

University of California, Santa Barbara

University of Stockholm

Edith Moravcsik

Gerrit Dimmendaal

University of Wisconsin, Milwaukee

University of Cologne

Masayoshi Shibatani

Martin Haspelmath

Rice University and Kobe University

Max Planck Institute For Evolutionary Anthropology, Leipzig

Russell Tomlin

Brian MacWhinney

University of Oregon

Volume 70 Issues in the Study of Pidgin and Creole Languages by Claire Lefebvre

Issues in the Study of Pidgin and Creole Languages

Claire Lefebvre Université du Québec à Montréal and Max Planck Institute for Evolutionary Anthropology, Leipzig

John Benjamins Publishing Company Amsterdam/Philadelphia

8

TM

The paper used in this publication meets the minimum requirements of American National Standard for Information Sciences – Permanence of Paper for Printed Library Materials, ansi z39.48-1984.

Library of Congress Cataloging-in-Publication Data Claire Lefebvre Issues in the Study of Pidgin and Creole Languages / Claire Lefebvre. p. cm. (Studies in Language Companion Series, issn 0165–7763 ; v. 70) Includes bibliographical references and indexes. 1. Pidgin languages. 2. Creole dialects. 3. Creole dialects, French-Haiti. PM7802 .L44 2004 417/.22 22 isbn 90 272 3080 3 (Eur.) / 1 58811 516 X (US) (Hb; alk. paper)

2004041134

© 2004 – John Benjamins B.V. No part of this book may be reproduced in any form, by print, photoprint, microﬁlm, or any other means, without written permission from the publisher. John Benjamins Publishing Co. · P.O. Box 36224 · 1020 me Amsterdam · The Netherlands John Benjamins North America · P.O. Box 27519 · Philadelphia pa 19118-0519 · usa

CONTENTS List of tables Preface ; List of abbreviations CHAPTER 1 Introduction 1.1 The perspective of this book 1.2 Content of chapters 1.3 The use of the words pidgins and creoles CHAPTER 2 The genesis of pidgin and creole languages: A State of the Art 2.1 The complex problem of pidgin and creole genesis 2.2 Competing theories of pidgin and creole genesis whose object of study consists of language varieties 2.2.1 The theory of pidgins and creoles as reduced codes 2.2.2 The theory of creoles as 'nativised pidgins' 2.2.3 The theory of PCs as crystallised varieties of 'imperfect' second language acquisition 2.2.4 The theory of PCs as restructured varieties 2.2.4.1. PCs as restructured substratum varieties 2.2.4.2. PCs as restructured superstratum varieties 2.2.4.3. PCs as restructured varieties of both of their source languages .... 2.2.5 The theory that creoles reflect the properties of Universal Grammar 2.2.6 Summary 2.3 Shifting the object of study from language varieties to processes involved in language creation and change 2.3.1 Relexification 2.3.2 Reanalysis and related phenomena 2.3.3 Dialect levelling 2.3.4 Summary 2.4 A unified theory of pidgin and creole genesis 2.4.1 Hypothesis and methodology of the UQAM projects 2.4.2 The interplay of the processes embedded within a scenario of creole genesis 2.4.3 An optimal account of creole genesis 2.4.4 Summary 2.5 Conclusion

xi xiii xv

1 1 2 5

7 8 12 12 14 16 17 17 17 18 19 21 23 24 26 27 28 28 29 31 33 34 34

vi

ISSUES IN THE STUDY OF PIDGIN AND CREOLE LANGUAGES

CHAPTER 3 The relexification account of creole genesis. The case of Haitian Creole. 3.1 The relexification hypothesis of creole genesis 3.1.1 The process of relexification 3.1.2 The hypothesis 3.1.3 The interplay of relexification, reanalysis and dialect levelling in creole genesis and development 3.1.4 Word order 3.1.5 An optimal account of creole genesis 3.1.6 Summary 3.2 The test of the hypothesis 3.2.1 The historical research 3.2.2 The linguistic test 3.3 The lexicon 3.3.1 Lexical semantics 3.3.1.1 Simplex and compound nouns 3.3.1.2 Pronouns 3.3.1.2.1 Personal pronouns 3.3.1.2.2 Possessive adjectives and pronouns 3.3.1.2.3 Logophoric pronouns 3.3.1.2.4 Expletives 3.3.1.3 Reflexives 3.3.1.4 Wh-words 3.3.1.5 Verbs 3.3.1.6 Summary 3.3.2 The syntactic properties of verbs 3.3.2.1 Types of argument structures 3.3.2.2 BODY-state verbs 3.3.2.3 WEATHER verbs 3.3.2.4 Reflexive verbs 3.3.2.5 Verbs licensing expletive subjects 3.3.2.6 Raising verbs 3.3.2.7 Existential verbs 3.3.2.8 Control verbs 3.3.2.9 Light verbs 3.3.2.10 Inherent object verbs 3.3.2.11 The case-assigning properties of verbs 3.3.2.12 Double-object verbs 3.3.2.13 Summary 3.3.3 Derivational affixes 3.3.4 Functional category lexical entries involved in nominal structure 3.3.4.1 The definite determiner 3.3.4.2 The plural marker 3.3.4.3 The indefinite determiner 3.3.4.4 The deictic terms 3.3.4.5 Case markers within the noun phrase 3.3.4.6 Summary 3.3.5 Functional category lexical entries involved in clause structure

37 38 38 42 43 46 46 47 48 48 48 49 49 50 52 52 54 56 57 58 62 67 68 69 69 71 72 74 74 75 77 77 80 81 83 83 84 84 88 89 90 92 92 94 94 95

CONTENTS 3.3.5.1 3.3.5.2 3.3.5.3

3.4

3.5

The tense, mood and aspect markers Complementisers and complementiser-like forms Complementisers or resumptives in the context of extracted subjects? 3.3.5.4 The nominal operator in relative and factive clauses 3.3.5.5 Clausal conjunction 3.3.5.6 The cleft marker 3.3.5.7 Negation markers 3.3.5.8 Markers expressing the speaker's point of view 3.3.5.9 The determiner in the clause 3.3.5.10 Summary Parameters 3.4.1 The null subject parameter 3.4.2 Verb raising 3.4.3 Serial verbs 3.4.4 The double-object construction 3.4.5 The interpretation of negative quantifiers 3.4.6 Verb-doubling phenomena 3.4.7 Summary Conclusion and consequences

CHAPTER 4 What do creole studies have to offer to mainstream linguistics? 4.1 In what sense do creole studies constitute a field? 4.2 What do creole studies have to offer to linguistics? 4.3 What progress has been made between 1994 and 1999? 4.4 What drawbacks have been overcome? 4.4.1 The Tower of Babel 4.4.2 Isolating Atlantic creoles from Pacific creoles 4.4.3 Isolating the study of pidgin and creole languages from that of mixed languages 4.4.4 Setting issues outside the research paradigms of subdisciplines of linguistics 4.5 What have creole studies contributed so far and what lies ahead?

vii 95 98 102 102 104 105 106 108 108 109 110 110 111 112 113 114 116 120 120 125 125 127 130 133 133 134 135 135 136

CHAPTER 5 On data 5.1 The non-neutral character of linguistic data 5.2 How to overcome the limits of particular types of databases 5.3 The problem of "inconsistencies" in elicited data 5.3.1 The problem of inconsistencies between speakers 5.3.2 The problem of inconsistencies within a given speaker 5.4 Native speakers as informants 5.5 Can data on a given language ever be complete?

139 140 143 148 148 152 153 154

CHAPTER 6 Multifunctionality and the concept of lexical entry 6.1 Establishing the various functions of multifunctional lexical items 6.2 The range of multifunctional lexical items across syntactic categories

155 157 161

viii 6.3 6.4

6.5 6.6 6.7 6.8 6.9

ISSUES IN THE STUDY OF PIDGIN AND CREOLE LANGUAGES Multifunctionality, the monosemy principle and semantic underspecification Categorial underspecification 6.4.1 Theoretical assumptions 6.4.2 The underspecification of major category lexical items 6.4.3 The underspecification of functional category lexical items An underspecification account of Fongbe 'to say' : verb and complementiser On the relationship between multifunctionality and grammaticalisation Grammaticalisation and multifunctionality are not synonyms Apparent cases of grammaticalisation in creole languages The transfer of substratum multifunctional lexical items into a creole

165 168 168 168 171 173 175 176 178 180

CHAPTER 7 On the semantic opacity of creole languages 7.1 Lexicon 7.1.1 Lexical semantics 7.1.2 Idiomatic expressions 7.1.3 Paradigms of functional categories 7.1.4 Phonologically null forms 7.1.5 Multifunctionality or cases of categorial neutralisation 7.1.6 Summary 7.2 Morphology 7.2.1 Suffixes that refer to a place of origin 7.2.2 The prefix ti7.2.3 The prefixde7.2.4 Summary 7.3 Syntax: basic word order and movement rules 7.4 Interpretive facts 7.4.1 The interpretation of tense, mood and aspect 7.4.2 The interpretation of clauses involving argument alternations 7.4.3 Verb doubling phenomena 7.4.4 The interpretation of cleft constructions 7.4.5 Summary 7.5 Language specific versus language universal phenomena 7.6 Global evaluation of the Semantic Transparency Hypothesis

181 183 183 184 185 185 186 186 188 188 188 190 196 196 198 199 200 201 202 203 203 204

CHAPTER 8 Do creole languages really form a typological class? 8.1 What is similar among creole languages? 8.2 Why do creole languages tend to be isolating? 8.3 Why do creole languages look simple? 8.4 Apparent simplicity and hidden complexity 8.5 McWhorter's list revisited 8.6 Conclusion

207 207 217 220 223 225 228

CHAPTER 9 The interplay of relexification and levelling in creole genesis and development 9.1 Relexification and levelling 9.2 The linguistic situation in Haiti at the time Haitian Creole was formed 9.3 Third person plural pronouns and plural markers

231 233 238 241

CONTENTS 9.4 9.5

9.6

Reflexives Demonstrative terms 9.5.1 The Haitian lexicon that has both sa and sila 9.5.2 The Haitian lexicon that has sa but not sila 9.5.3 Levelling Conclusion

CHAPTER 10 The emergence of productive morphology in creole languages: The case of Haitian Creole 10.1 Identifying affixes that are potentially native to a creole language 10.2 Evaluating the productivity of the affixes identified as potentially native to HC 10.2.1 The agentive suffix -è 10.2.2 The attributive suffix-è 10.2.3 The verbalising suffix -e 10.2.4 Inversive and privative de10.2.5 The diminutive prefix ti10.2.6 The nominalising suffix -ay 10.2.7 Morphological conversion 10.2.8 The adverbial suffix -man 10.2.9 The place of origin/residence suffixes -wa and -yen 10.2.10 The ordinal suffix -yèm 10.2.11 Summary 10.3 Evaluating other proposals in the literature 10.3.1 The hypothesised inversive prefix en10.3.2 The hypothesised suffix -et 10.3.3 The hypothesised nominalising suffix -man 10.3.4 The hypothesised agentive suffix -add 10.3.5 Summary 10.4 The emergence of the morphological inventory of HC 10.4.1 The inventory of the productive derivational affixes of HC as compared with those of its contributing languages 10.4.2 The relexification account of the emergence of the HC morphological inventory 10.4.3 Summary 10.5 Conclusion: Issues on the morphology of creole languages in light of the HC data..

ix 245 250 250 255 256 257

259 260 262 263 263 264 264 266 268 268 270 271 271 272 273 273 275 279 286 292 292 293 300 303 303

References

309

Appendices Appendix I: A research programme on PC genesis for the 21st century Appendix II: The composition of the Haitian lexicon

341 345

Index of authors Index of subjects

349 355

LIST OF TABLES 2.1 Summary of the six theories of pidgin and creole genesis with respect to the features that need to be accounted for by any theory that seeks to account for the origin of these languages

22

3.1 The inventory of HC productive affixes 3.2 The HC affixes and their closest French corresponding forms

86

3.3 The HC affixes and their Fongbe corresponding forms

87

3.4 The inventory of TMA markers in Haitian and in Fongbe 4.1 Number of papers per topic in JPCL (1994-1999) 4.2 Number of papers per topic in the "Westminster Creolistics Series" (volumes 1 and 2) 4.3 Number of papers per language (discussed in more than one paper) 8.1 The inventory of TMA markers in Haitian and in Fongbe 8.2 Comparison of the parametric options in the three languages under comparison

85

96 130 131 132 211 212

10.1 The inventory of HC productive affixes

273

10.2 The HC affixes and their closest French corresponding forms 10.3 The HC affixes and their Fongbe corresponding forms

295 297

PREFACE The content of this book is concerned with various issues at stake in creole studies that are also of interest for general linguistics. This book is conceived of as a reader on topics as varied as cognitive and social processes involved in linguistic change, the problem of the validity of linguistic data, multifunctionality, language typology, etc. With the exception of chapter 1, the introduction to the book, and chapter 3, written specifically for this publication, the chapters of this book consist of a subset of the articles that I have written since the publication of my book by Cambridge University Press in 1998: Creole Genesis and the Acquisition of Grammar: The Case of Haitian Creole. These articles all explore topics that were not fully developed in my book. These articles were originally published in various journals and books. Chapter 2 "The Genesis of Pidgin and Creole Languages: A State of the Art" appeared in 2003 as part of a collection of papers edited by C. Jourdan and K. Tuite, entitled Ethnolinguistics and Anthropological Theory, published by Fides. The Appendices to chapters 2 and 3 consist of sections of "The Field of Pidgin and Creole Linguistics at the Turn of the Millennium: The Problem of the Genesis and Development of PCs" published in 2002 as part of a collection of papers edited by G. Gilbert, Pidgin and Creole Linguistics in the Twenty-First Century, published by Peter Lang (pp. 247-287). Chapters 4, 5, 6 and 7 consist of four essays that I wrote as the columnist for the Journal of Pidgin and Creole Languages during the years 2000 and 2001. Thus, the content of chapter 4 appeared in 2000 as "What do Creole Studies have to Offer to Mainstream Linguistics?" in Journal of Pidgin and Creole Languages 15,1: 127-155; the content of chapter 5 was also published in 2000 under the title "On Data", Journal of Pidgin and Creole Languages 15,2: 313-337; the content of chapter 6 "Multifunctionality and the Concept of Lexical Entry" was published in 2001 in Journal of Pidgin and Creole Languages 16,1: 107-145, and that of chapter 7 "On the Semantic Opacity of Creole Languages" was also published in 2001 in Journal of Pidgin and Creole Languages 16,2: 321-353. The content of chapter 8 was published in 2001 under the title "What You See Is Not Always What You Get: Apparent Simplicity and Hidden Complexity in Creole Languages" in Linguistic Typology 5,2/3: 186-213. The content of chapter 9 appeared in 2001 under the title 'The Interplay of Relexification and Levelling in Creole Genesis

xiv

ISSUES IN THE STUDY OF PIDGIN AND CREOLE LANGUAGES

and Development" in Linguistics 39,2: 371-408. The content of chapter 10 consists of an extended version of "The Emergence of Productive Morphology in Creole Languages: The Case of Haitian Creole" that appeared in 2003 in the Yearbook of Morphology 2002 edited by G. Booij and J. van Marie (pp. 35-81). I would like to thank the editors and the publishers of all the sources mentioned above for giving me permission to reproduce in part or in whole the papers that I originally published with them. Several scholars gave me comments on drafts of the various papers that constitute chapters of this book. They are all acknowledged in the first note of each chapter. The research underlying the content of this book was carried out within the context of various funded research projects. These were mainly funded by SSHRCC with complementary grants from FCAR and FIR-UQAM. The preparation of the final manuscript of this book was financed by SSHRCC, the Max Planck Institute for Evolutionary Anthropology in Leipzig and FIR-UQAM. I thank these institutions for making the research reported on in this book and its publication possible. I want to thank my university and my department for supporting this research. My gratitude goes to Bernard Comrie for his invaluable support in making the results of this research public. I am indepted to my colleague Jean-François Prunet for encouraging me to put this book together. I also want to thank Bernard Comrie and Elly van Gelderen for their comments on the first version of the manuscript. Their comments contributed to the cohesion of the final version. My gratitude goes to Lisa Travis, Werner Abraham, Michael Noonan and Kees Vaes who made this publication possible. Finally, many thanks to Andrée Bélanger for preparing the camera-ready copy of this book, and to Anne-Sophie Bally, Maribel Olguin and Isabelle Themen for their help in preparing the indices.

LIST OF ABBREVIATIONS ACC accusative AG agentive affix ANT marker of anteriority ASP aspect AUX auxiliary COMP complementiser CONJ conjunction pl DEF definite determiner DEF.FUT definite future marker DEIC deictic term DEM demonstrative determiner DP determiner phrase f feminine FP focal pronoun FUT future GEN genitive case marker HAB habitual marker

Neg NEG NP OBJ OP POST PL POSS PROX Q REF RES SEL SELF SEM sg

negative marker negation marker nominal phrase objective case operator postposition plural plural marker possessive proximate question marker reflexive resumptive selectional properties anaphor semantic properties singular

IMP

imperfective aspect

SRP

subject-referencing pronoun

HMD IND.FUT INFL INS IRR LOC LOG

indefinite determiner indefinite future marker inflection marker of insistence irrealis mood marker locative logophoric pronoun

SUB SYN TMA TO VAL

subjunctive syntactic features tense, mood and aspect markers topic marker validator

MO

mood

CHAPTER 1

INTRODUCTION The content of this book is concerned with various issues at stake in creole studies that are also of interest for general linguistics. These include the general problem of creole genesis and of the accelerated linguistic change that characterises the emergence of these languages as compared to ordinary cases of linguistic change (chapters 2, 3, 4 and 10), the problem of the validity of data in linguistic analysis (chapter 5), the issue of multifunctionality as regards the concept of lexical entry (chapter 6), the question of whether creole languages are semantically more transparent than languages not known as creoles (chapter 7), the question of whether creole languages constitute a typologically identifiable class (chapter 8) and the problem of the interaction between the processes involved in the emergence and development of creole languages (chapter 9). 1 . 1 . The perspective of this book All the chapters in this book address the various issues from a unified perspective. First, the general perspective that links the chapters is centred around the tension that exists between the notion of grammar and that of language. This is because, as will be seen in the various chapters of this book, languages that are different may share the same grammar. This is the case, for example, of the West African Gbe languages, which constitute different languages due to the fact that they have different lexicons, but which to a great extent have similar grammars. The fact that the Gbe languages have similar grammars is not surprising since these languages constitute a dialect cluster (see Capo 1984). However, the fact that these languages and Haitian Creole have very similar grammars is not easily foreseen. The reverse of this situation may also be true. For example, while the lexicons of Haitian Creole and French are similar in their labels, the properties of their lexical items are quite distinct. While French exhibits the characteristics of Romance languages, Haitian Creole manifests those that are typical of Gbe (and more generally of Kwa) languages. As is evidenced by the data presented throughout this book, while the phonological representations of Haitian Creole lexical entries are derived from French phonetic matrices, the semantic and syntactic properties of these lexical entries are derived from the West African substratum languages of the creole.

2

ISSUES IN THE STUDY OF PIDGIN AND CREOLE LANGUAGES

Another aspect that links the chapters of this book is that all the issues concerning creole genesis are addressed from the point of view of the processes otherwise known to play a role in language genesis and language change in general. These processes are relexification, reanalysis and dialect levelling. Relexification essentially alludes to the assigning of new labels to the lexical entries of a given lexicon. The lexical entries so formed have the semantic and syntactic properties of the original entries with labels drawn from another language. This process has been shown to play a role in the formation of mixed languages such as Media Lengua (e.g. Muysken 1981a), Michif (e.g. Bakker 1992), Ma'a (e.g. Mous 1994), and of creole languages (e.g. Lefebvre 1998a, and the references therein). Reanalysis is a process whereby a particular form which signals one lexical entry becomes the signal of another lexical entry. Reanalysis is a major process in linguistic change (e.g. Hopper and Traugott 1993). Dialect levelling refers to the reduction of variation between dialects of the same language in situations where these dialects are brought together (e.g. Trudgill 1986). A third element that links the articles in this book is that all the issues are addressed on the basis of data drawn from Haitian Creole. Haitian is an important creole language due to both the numerical importance of its speakers (more than 7 million speakers, according to Grimes (ed.) 2000), and the extent to which it has been described and analysed in comparison to other creole languages. The various topics proposed for discussion thus provide a number of angles from which the Haitian Creole data can be looked at. In putting the articles together as a collection, I was concerned with preserving the integrity and the independent character of each article. Minor cuts have been made however, so as to prevent unnecessary repetitions. Crossreferences to chapters of this book have been incorporated, and minor clarifications have been added. The additions in the text all appear within square brackets. In some cases, I have added appendices. 1.2. Content of chapters The most intriguing question about pidgin and creole languages is no doubt that of how they come about. The problem of pidgin and creole genesis has been addressed from different points of view over the last century. Chapter 2 entitled "The Genesis of Pidgin and Creole Languages: A State of the Art" summarises and evaluates the various theories proposed to account for the genesis of these languages. The chapter begins with a discussion of features that any theory aiming at explaining the origin of pidgin and creole languages must be able to account for. Competing theories of creole genesis are presented and evaluated against these features. The conclusion to the chapter provides references identifying avenues for future research on this topic. An appendix entitled "A

INTRODUCTION

3

Research Programme on Pidgin and Creole Genesis for the 21 st Century" provides a list of major research topics to be addressed on the genesis issue. The global results of 25 years of research on creole genesis that I directed with the collaboration of colleagues at UQAM, based on an in-depth study of Haitian Creole and its contributing languages, have been published by Cambridge University Press in a 460 page book entitled Creole Genesis and the Acquisition of Grammar: The Case of Haitian Creole. Chapter 3 "The Relexification Account of Creole Genesis: The Case of Haitian Creole" consists of a summary of this book. The chapter begins with a summary of the theory of creole genesis developed in the course of this research. It summarises the data and analyses presented in the book. It is shown that the process of relexification plays a major role in the formation of Pidgins and Creoles and that two other processes, reanalysis and dialect levelling, applying on the output of relexification, also play a role in the makeup of these languages. The chapter ends with a discussion on the consequences of these findings. The field of creole studies has a clear contribution to make to general linguistics were it only for the question of how pidgins and creoles come about. Chapter 4 "What do Creole Studies have to Offer to Linguistics" is an essay on the relationship between creole studies and mainstream linguistics. It addresses the following questions: Do creole studies constitute a field? What do creole studies have to offer to linguistics? What progress was made between 1994 and 1999? What drawbacks were overcome? What have creole studies contributed so far and what lies ahead? Topics for future research on the basis of this overview are also identified. The problem of the validity of data is a permanent issue in general linguitics. Creolistics is no exception. Chapter 5, "On Data", is an essay on the issue of what constitutes valid data for linguistic analysis. The discussion is based on some thirty years of experience in data collection and analysis that I carried out on Quechua, French, Martinican Creole, Haitian Creole and Fongbe. Three major themes are addressed: the non-neutral character of linguistic data, how to overcome the limits of particular databases, and the problem of "inconsistencies" in elicited data. Although cases of multifunctionality may be found in all human languages, the number and range of multifunctional lexical items in creole languages has been claimed to be quite significant. Chapter 6 deals with "Multifunctionality and the concept of lexical entry". Because it is at the intersection of lexicon, semantics and syntax, the phenomenon of multifunctionality raises a number of theoretical questions such as the following: How can multifunctionality be characterised? What is an optimal account of the phenomenon? Furthermore, multifunctionality and the process of

4

ISSUES IN THE STUDY OF PIDGIN AND CREOLE LANGUAGES

grammaticalisation are not unrelated, for grammaticalisation may be viewed as a process creating multifunctional lexical items. The Semantic Transparency Hypothesis claims that creole languages are more transparent than languages not known as creoles. Chapter 7 entitled "On the Semantic Opacity of Creole Languages" challenges this hypothesis. The chapter begins with a review of the various facets of this hypothesis as formulated in Wekker (1982) and in Seuren and Wekker (1986), and of the predictions that it makes on the form of creole languages. The test of the Hypothesis comes down to two questions. First, considering the various components of the grammar (the lexicon, the morphology, the syntax and the semantic component), is there a one-to-one correspondence between form and meaning at different levels of representation? Second, if creoles are transparent, the non-transparent features of their contributing languages are expected to be eliminated in creolisation. Is this the case? The chapter points out all the possible ways in which a creole language can be considered opaque as opposed to transparent. It is shown that opacity may manifest itself in a creole at all levels of representation. The data presented in this chapter support the view that creole languages are no more transparent than languages not known as creoles, and that in some cases, they are even more opaque than non-creole languages. Data showing that the non-transparent features of the substratum languages are not necessarily suppressed in creolisation are also presented. In recent literature, it has been claimed that creole languages constitute a typologically identifiable class. In chapter 8, entitled "Do Creole Languages Really Form a Typological Class", I take issue with this claim and more particularly with McWhorter's (1998a, 2001) position that "creole grammars constitute a synchronically identifiable class". The following topics are discussed: What do creole languages really have in common? Why do creole languages tend to be isolating? Why do creole languages tend to look simple? These questions are addressed from the point of view of the relexification account of creole genesis in Lefebvre (1998a, and the references therein). The last section considers McWhorter's hypothesised creole typological features in light of the previous discussion. It is argued that the features proposed by McWhorter as identifying creole languages are derivable from a sound theory of how creole languages come about. Among the current issues surrounding the question of the development of creole languages figures the question of how the process of dialect levelling interacts with the process of relexification. Chapter 9, entitled "The Interplay of Relexification and Levelling in Creole Genesis and Development" takes up this issue. It is a well documented fact that situations where creoles develop involve several substratum languages and one (or at least one major) superstratum

INTRODUCTION

5

language. The relexification of several substratum languages on the basis of a single/major superstratum language provides the creole community with a common lexicon, hence a common language. Being a mental/cognitive process, relexification is an individual process. Consequently, through relexification, the specificities of the various substratum lexicons are reproduced in the incipient creole, thus creating what might be referred to as different "dialects" of the new language. The relexification of several substratum languages on the basis of a single superstratum language can be viewed as the major source of variation in an incipient creole. This provides a sound explanation for the fact that different substratum languages may contribute different features to a particular creole. Dialect levelling, as discussed in the literature on dialects in contact, refers to the reduction of variation between dialects of the same language, in situations where these dialects are brought together. On the basis of three sets of data from Haitian Creole, it is shown that in a Creole's development, dialect levelling operates on the output of relexification. The role of levelling in creole genesis and development accounts for the fact that the properties of some specific lexical entries of the creole may depart from those of the corresponding lexical entries in the individual substratum languages. It is a well known fact that creole languages derive the bulk of their vocabulary from their superstratum language. However, a word that has internal structure in the superstratum language may not have internal structure in the creole corresponding word. Criteria are needed in order to determine whether a given creole word has internal structure and what affixes are native to and productive in this language. Chapter 10 entitled "The Emergence of Productive Morphology in Creole Languages: The Case of Haitian Creole" addresses issues concerning the methodologies that need to be developed for studying the emerging morphology of creole languages. One major issue gravitates around the criteria for identifying native and productive affixes in a given creole. Another concerns the account of the emergence of productive morphology in a creole. A number of potential Haitian affixes will be considered and criteria will be proposed to evaluate their nativeness and productivity. The content of this chapter takes issue with, among other authors, DeGraff (2001) on the range of affixes that are productive in Haitian Creole. 1 . 3 . The use of the words pidgins and creoles Pidgins and creoles have long been considered as separate entities on the basis of the following two sets of criteria. While pidgins have been defined as reduced codes, creoles have often been defined as expanded versions of these reduced codes (e.g. Hymes (ed.) 1971). Also, while pidgins have been found to constitute the second language of the speakers who use them, a creole is often considered to be a pidgin that has become the first language of a new generation

6

ISSUES IN THE STUDY OF PIDGIN AND CREOLE LANGUAGES

of speakers (e.g. Kay and Sankoff 1974). In more recent literature, the distinction between pidgins and creoles has been levelled out in view of the fact that there are some pidgins (still used as a second language) that have been shown to have expanded in the same way as languages known as creoles (e.g. Mühlhäusler 1980, 1986a, for an extensive discussion of this point). Hancock (1980: 64) states: "I prefer not to acknowledge a distinction between pidgin and creole, and to consider stabilisation more significant than nativisation in creole language formation." Similarly, Mufwene (1990: 2) uses the term creole to refer "to varieties traditionally called creoles but also to those called pidgins that serve as vernaculars or primary means of communication for at least a portion of their speakers." Moreover, in recent literature in the field, scholars have started referring to pidgins and creoles as PCs, suggesting that they fall into a single category. Furthermore, as is pointed out in Lefebvre (1998a, chapter 2), pidgin and creole languages cannot be distinguished on the basis of the processes involved in their formation (see also Woolford 1983, for a general discussion of this point). Indeed, the processes hypothesised to play a role in the formation and development of human languages apply to both pidgins and creoles. For these reasons, in this book, I do not distinguish between pidgins and creoles, and I sometimes refer to them as PCs.

CHAPTER 2

THE GENESIS OF PIDGIN AND CREOLE LANGUAGES A STATE OF THE ART*

The most intriguing question about pidgin and creole languages is no doubt that of how they come about. This question is also most important for it hinges on several disciplines, as will be shown throughout this chapter. Due to the circumstances in which they develop, pidgins and creoles constitute an extreme case of languages in contact. Their creation involves second language acquisition, and their development, first language acquisition. The emergence of pidgin and creole languages constitutes a particular case of accelerated linguistic change; it is thus a goldmine for historical linguistics. Two major processes involved in the creation and development of these languages, relexification and reanalysis, are cognitive processes; hence, the study of the emergence of pidgin and creole languages is relevant to cognitive sciences. The other process that plays a major role in the development of pidgins and creoles, levelling, is a social process; hence, the study of the emergence of pidgin and creole languages is of interest to socio- and ethno-linguistics. This paper is concerned with the various linguistic dimensions of the emergence of pidgin and creole languages. The problem of the origin of pidgin and creole genesis has been addressed from different points of view over the last century. This chapter summarises and evaluates the various theories proposed to account for the genesis of these languages. The chapter begins with a discussion of features that any theory aiming at explaining the origin of pidgin and creole languages must be able to account for (section 2.1). The next three sections evaluate competing theories of creole genesis against these features. Section 2.2 reviews theories that * The paper reproduced as the content of this chapter first appeared in a collection put together by Christine Jourdan and Kevin Tuite (2003) (Lefebvre 2003a). The appendix to this chapter, Appendix 1, entitled "A research programme on PC genesis for the 21st century" consists of a section of a larger paper, "The field of pidgin and creole linguistics at the turn of the millennium: The problem of the genesis and development of PCs", prepared for the symposium on Pidgin and Creole Linguistics in the 21st Century held in New York (January 1998) as part of the annual Linguistic Society of America meeting, and published in Gilbert (2002) (Lefebvre 2002). This chapter has been written as part of the project "Research builds on research" financed by SSHRCC. I want to thank the following colleagues for their most useful comments on an earlier version of this paper: Bernard Comrie, Christine Jourdan, Jeff Siegel and Kevin Tuite. Thanks to Marijo Denis and Virginie Loranger for their help in editing the manuscript, and to Andrée Bélanger for formatting it.

8

ISSUES IN THE STUDY OF PIDGIN AND CREOLE LANGUAGES

share the property of having language varieties as an object of study. Section 2.3 presents the first attempts to shift the focus of pidgin and creole genesis studies from the language varieties to the processes involved in the formation of these languages. Section 2.4 reports on a major long term research project, carried out at the Université du Québec à Montréal, that aimed at providing a unified account of pidgin and creole genesis within the framework of the processes that otherwise play a role in language creation and language change in general. Section 2.5 concludes the chapter and provides references identifying avenues for future research on the topic. Appendix 1 provides additional discussion of topics for future research. 2.1. The complex problem of pidgin and creole genesis The history and structure of PCs are characterised by a number of features.1 Any theory that seeks to explain the origin of these languages must be able to account for this basic set of features. First, as was pointed out by Whinnom (1971) these languages are only developed in multilingual communities. Whinnom argues that, in bilingual communities, the speakers of one group will eventually learn the language of the other group.2 Second, communities where PCs emerge generally involve several substratum languages whose speakers make up the majority of the population and a superstratum language spoken by a relatively small but economically powerful social group. Crucially, the substratum community does not have one common language. This situation creates the need for a lingua franca (see e.g. Foley 1988; Hymes 1971; etc.), not only to permit communication between the speakers of the substratum languages and of the superstratum language, but also to permit the speakers of the substratum languages to communicate among themselves (see e.g. Foley 1988; Singler 1988: 47; Thomason and Kaufman 1991). Third, in communities where PCs emerge, speakers of the substratum languages generally have very little access to the superstratum language (see Thomason and Kaufman 1991). As Foley (1988: 163) puts it: "the language of 1 The content of this section builds on a preliminary discussion in Lefebvre and Lumsden (1989, 1994a) updated in Lefebvre (1998a: 1-4). 2 There is a general consensus in the literature that multilingualism is a required feature of communities where creoles may emerge. However, this proposal has recently been challenged by Smith, Robertson and Williamson (1987), who claim that Berbice Dutch emerged out of contact between only two languages: Dutch and Eastern Ijo. Assuming that Berbice Dutch is a true creole, and that Ijo was the sole African language present at the time this creole was formed, this case would constitute the first documented evidence against Whinnom's widely accepted claim. For further discussion of this issue, see also Foley (1988) and Thomason (1997b).

THE GENESIS OF PIDGIN AND CREOLE LANGUAGES

9

the dominant group is not easily made available to the members of the subordinate group(s)." In fact, as has been pointed out on several occasions in the literature, creoles that most resemble their superstratum languages were created in communities where the speakers of the substratum languages had relatively more access to the superstratum community. Creoles that are more radical (i.e. less like the superstratum language) come from communities where language learners had very little access to the superstratum community (see e.g. Andersen 1983; Baker 1993; Baker and Corne 1982; Bickerton 1977: 55; Thomason and Kaufman 1991; Valdman 1978, 1993). For example, as is argued in Valdman (1993), Louisiana Creole is closer to French than Haitian is because the substratum speakers had more access to French in Louisiana than the African population had in Haiti. Baker and Corne (1982) also discuss this issue on the basis of data from Mauritius and Reunion creoles. On Reunion, French native speakers outnumbered substratum speakers during the formative period of the creole and Reunion Creole grammar displays a significant number of French grammatical categories. By contrast, during the formative period of Mauritius Creole, the proportion of native French speakers was much lower and thus the West African speakers had a much stronger input into the creole. Likewise, during the period where Haitian Creole was formed, the proportion of West African speakers was much higher than that of French speakers (see Singler 1996), such that West African speakers had a very strong input into the creole (see Lefebvre 1998a, and the references cited therein). A fourth point is that, ordinarily, languages change gradually. Within the span of several generations, speakers of innovative and conservative dialects are able to communicate, even though, over the course of centuries, a new language may evolve (see Lightfoot 1979). By contrast, PCs are created in a relatively short span of time (see e.g. Alleyne 1966; Bickerton 1984; Chaudenson 1977, 1993; Hall 1958; Voorhoeve 1973). This observation dates back to Van Name (1869-70: 123, cited in Goodman 1964: 135): "Under ordinary conditions these changes proceed at so slow a pace as to be appreciable only at considerable periods of time, but here two or three generations have sufficed for a complete transformation." Hesseling (1933: xi) further reassesses this point in the following terms: The genesis of human language is a psychological problem that no single language will ever solve, but from creole one can best learn how a given language emerges from old data and develops, because here something takes shape at a high speed, in a past recognisable to us, something which is the product, in other cases, of many centuries, with a very obscure past in its background.

10

ISSUES IN THE STUDY OF PIDGIN AND CREOLE LANGUAGES

Thus, in contrast to regular cases of linguistic change, PCs diverge abruptly from their source languages (see Thomason and Kaufman 1991), such that within one or two generations, a different language is created. Hancock (1987: 265) claims that: "most of the principal characteristics that each creole is now associated with were established during the first twenty-five years or so of the settlement of the region in which it came to be spoken." Ferraz (1983), Hymes (1971) and Mintz (1971) suggest that a creole can develop within fifty years or less. Singler (1996) is of the opinion that it takes sixty to eighty years for a creole to form. Whatever the outcome of this issue may be, PCs constitute a unique case of accelerated linguistic change when compared with regular cases of linguistic change. Fifth, PCs tend to be isolating languages. This observation goes back to Hesseling (1933: xvi) and Schuchardt (1979). It is also found in Hagège (1985: 39). But it was Mufwene (1986, 1990, 1991) who clearly established this property of PCs and the problem it poses for scholars who work on PC genesis. Indeed, Mufwene has documented the fact that this tendency appears to hold even when the contributing languages are not isolating languages. For example, Mufwene (1986) shows that Kituba, a creole language that has emerged almost exclusively from contact among agglutinative Bantu languages, is an isolating language. "Kituba has selected Kikongo's seemingly marked periphrastic alternative over the more common and apparently unmarked agglutinating system" (Mufwene 1990: 12). Sixth, it has long been noted in the literature that PCs derive some of their properties from those of the substratum languages and some from those of the superstratum language (see e.g. Alleyne 1966, 1980; Holm 1988; etc.). Moreover, several scholars have noticed that the respective contribution of the substratum and superstratum languages to a creole is not random. For example, Adam (1883: 47) states that: J'ose avancer... que les soi-disant patois de la Guyane et de la Trinidad constituent des dialectes négro-aryens. J'entends par là que les nègres guinéens, transportés dans ces colonies, ont pris au français ses mots, mais qu'ayant conservé dans la mesure du possible, leur phonétique et leur grammaire maternelles... Une telle formation est à coup sûr hybride... La grammaire n'est autre que la grammaire générale des langues de la Guinée. [I go so far as to claim... that the so-called patois of Guyana and Trinidad constitute Negro-Aryan dialects. By that I mean that the Guinean Negroes who were transported to the colonies adopted the words of French but, as much as possible, kept the phonetics and grammar of their mother tongues... Such a formation is clearly hybrid... The grammar is no different from the general grammar of the languages of Guinea.]

THE GENESIS OF PIDGIN AND CREOLE LANGUAGES

11

Speaking of Haitian Creole, Sylvain (1936: 178) observes that: "Nous sommes en présence d'un français coulé dans le moule de la syntaxe africaine, ou (...) d'une langue éwé à vocabulaire français." [We are in the presence of a French that has been cast in the mould of African syntax or ... of an Ewe language with a French vocabulary.] Similarly, in his extensive study of French-based creoles, Goodman (1964) observes, over and over again, that particular lexical items in the creoles have a phonological representation similar to a French expression but that these creole lexical items share properties with corresponding lexical items in the African substratum languages. On the basis of data drawn from Djuka, Hurtar (1975: 684) also remarks that "the use of morphemes borrowed by a pidgin or a creole language (...) from a European language often diverges from the use of the source morpheme in the source language" and often corresponds to the use of the corresponding word in the substratum languages. Voorhoeve (1973) makes a similar remark on the basis of Sranan and Saramaccan data. Writing about Solomons Pidgin, Keesing (1988: 1-2) remarks: I had earlier been struck, when I had learned Solomons Pidgin in the 1960s through the medium of Kwaio, an indigenous language I already spoke fluently, that this learning task mainly required learning Pidgin equivalents of Kwaio morphemes. The syntax of Solomons Pidgin was essentially the same as the syntax of Kwaio, although somewhat simpler and lacking some of the surface marking; in most constructions, there was a virtual morpheme-by-morpheme correspondence between Kwaio and Pidgin. (...) Although most of the Pidgin lexical forms were ultimately derived from English, I found this largely irrelevant to my language-learning task. The semantic categories they labeled corresponded to Kwaio ones, not English ones; grammatical morphemes corresponded to Kwaio ones, not English ones. Thus semantically Pidgin dae corresponded directly to Kwaio mae 'be dead, die, be comatose, be extinguished,' not to English die. Pidgin baebae corresponded to the Kwaio marker of future/nonaccomplished mode, ta-, not to English by and by.

These observations suggest that PCs are not formed by an arbitrary mixture of the properties of the languages present at the time they are being created. The general pattern that seems to emerge from the observations reported above is the following: while the forms of the lexical entries of a PC tend to be derived from the superstratum language, the syntactic and semantic properties of these lexical entries tend to follow the pattern of the substratum languages. Any theory of PC genesis must account for the properties of these languages. Therefore, as has been pointed out in Lefebvre and Lumsden (1989), an optimal theory must account for the fact that PCs emerge in multilingual contexts, where there is a need for a lingua franca, and where the speakers of the substratum languages have little access to the superstratum language. It must account for the fact that PCs tend to be isolating languages even when they emerge from contact situations involving only agglutinative languages. It must

12

ISSUES IN THE STUDY OF PIDGIN AND CREOLE LANGUAGES

also account for the fact that PCs manifest properties of both their superstratum and substratum languages and it must explain why these properties are divided the way they are. Finally, an adequate theory of PC genesis must be stated in terms that are explicit enough so as to be falsifiable. 2.2.

Competing theories of pidgin and creole genesis whose object of study consists of language varieties This section summarises approaches to PC genesis that share the characteristic of having language varieties (as opposed to processes that lead to these varieties) as their object of study. The following proposals will be discussed in turn: The theory according to which PCs constitute reduced linguistic codes; the theory advocating that creole languages are nativised pidgins; the theory according to which creoles are imperfect second language varieties of their lexifier languages; the theories advocating that PCs consist of restructured varieties of their substrate and superstrate languages, respectively; and finally, the theory that creoles consist of language varieties reflecting the properties of Universal Grammar. (This last theory is referred to in the literature as the Language Bioprogram Hypothesis.) Following the methodology in Lefebvre and Lumsden (1989), each of these approaches will be evaluated on the basis of the set of features enumerated in section 2.1. A methodological proviso is in order at this point. I do believe that a sound theory of PC genesis must be able to account for all the seven features discussed in section 1. The theories that I evaluate on the basis of these features may, however, have not been formulated so as to account for this particular set of features. My evaluation of each theory is nevertheless based on whether the theory under review, in its current formulation, may account for the particular features in my list, regardless of whether the proposer meant to account for these or not. In adopting this methodology, I do not evaluate each theory for its own sake. However, I do provide the reader with a means of evaluating the various theories against a single set of criteria. 2.2.1 The theory ofpidgins and creoles as reduced codes This theory holds that speakers of the substratum languages were presented with a reduced (baby-talk or foreigner-talk) version of the superstratum language characterised by an absence of functional categories such as gender, number, case, etc. In this view, the plantation owners were voluntarily speaking a reduced version of their own language in order to maximise communication with the slave population. Bloomfield (1933), Göbl-Galdi (1934), Hall (1966), Jespersen (1922: 233) and Schuchardt (1909) (as translated by Goodman 1964: 124) all hold some version of this view. Similar proposals have also been made more recently. For example, Ferguson (1971: 147) advocates the view that "the

THE GENESIS OF PIDGIN AND CREOLE LANGUAGES

13

foreigner talk of a speech community may serve as an incipient pidgin. This view asserts that the initial source of the grammatical structure of a pidgin is the more or less systematic simplification of the lexical source language which occurs in the foreigner talk registers of its speakers, rather than the grammatical structure of the language(s) of the other users of the pidgin." (See Naro 1978, for an extensive discussion of this issue.) Similarly, Foley (1988: 166) writes: "I suggest that a pidgin is a version of a foreigner talk of a superstratum community that has been conventionalised and accepted, most importantly by speakers of the substrate language(s)." Does this approach meet the criteria of an adequate theory of PC genesis as outlined in section 2.1? This approach does not explain why PCs only develop in multilingual communities (the first feature). It does not explain the need for the substratum speakers to develop a lingua franca (the second feature). It does not account for the fact that these languages are formed quite rapidly (the fourth feature), nor for the type of mix that these languages manifest (the sixth feature). However, this approach may be considered to provide an account of the fact that the substratum population has little access to the superstratum language in situations where PCs emerge (the third feature), and of the lack of inflectional morphology, thus of the isolating character of these languages (the fifth feature). Finally, the theory of reduction is formulated in terms that are precise enough so as to be falsified (the seventh criterion). The next paragraph shows how this approach to PC genesis can be falsified and how it is in fact falsified. According to some of the proponents of this approach, creole languages would lack the functional category lexical entries of their superstratum language because the speakers of the substratum languages were not presented with these lexical items. This claim can be falsified if it can be shown that the speakers of the substratum languages of a creole were in fact presented with the pertinent data. As has been pointed out in Lefebvre (1998a: 62-65), in a discussion concerning the origin of Haitian Creole, while it could well be the case that French speakers did not use a very elaborate style of French while talking to the African population in Haiti, there is plenty of evidence from the Haitian lexicon that the speakers of the substratum languages were exposed to the functional category lexical items of French. Indeed, Valdman's et al. (1981) dictionary abounds in examples where a Haitian word corresponds to a French expression that includes a French functional item. As is shown in (1) (from Lefebvre 1998a: 64), a simple Haitian lexical entry may contain an agglutinated French determiner (e.g. la, au), as in (la), an agglutinated French partitive determiner (e.g. du), as in (lb), an agglutinated French complementiser (que 'that'), as in (lc), an agglutinated French functional item à, as in (ld), or even an agglutinated French conjunction (e.g. et, ou), as in (le).

14 (1)

ISSUES IN THE STUDY OF PIDGIN AND CREOLE LANGUAGES HAITIAN LEXICAL ENTRY FRENCH EXPRESSION

a. larivyè olye b. diri dife c. fok tandiske d. afòs apati e. epi oubyen

'river' 'instead' 'rice' 'fire' 'complementiser' 'while' 'by means of' 'starting from' 'and' 'or'

la rivière au lieu du riz du feu (il) faut que tandis que à force de à partir de et puis ou bien

'the river' 'instead' 'rice' 'fire' 'there must be' 'while' 'by means of 'starting from' 'and then' 'or'

These examples, and many more, show that the African population in Haiti must have been exposed to forms containing French functional items. My conclusion, based on Haitian, is in line with Alleyne (1971: 170), who states that the African population of the Caribbean area was exposed to European languages "in their full morphological and syntactic forms." The above data constitute a major drawback to the claim that a PC lacks the functional categories of its superstratum language because the substratum speakers were not exposed to these categories. Instead, they show very clearly that the creators of Haitian, and presumably of other PCs, were exposed to the superstratum functional categories, but that they did not identify them as such because they did not have enough exposure to the superstratum language. 2.2.2 The theory of creoles as 'nativised pidgins ' The idea that creole languages are nativised pidgins emerged during the late sixties and developed in the seventies. In this approach, a pidgin language is no one's first language. It is a reduced language variety serving as a lingua franca in a multilingual community. A pidgin that came to be spoken as the first language of a generation of speakers is said to have undergone nativisation. A nativised pidgin is called a creole. From a linguistic point of view, the nativisation of a pidgin is often seen as being accompanied by expansion or complexification of the source pidgin, the latter being claimed to acquire all the characteristics of a natural language in the process of nativisation (see e.g. Bickerton 1981; Hymes 1971; Labov 1971; Sankoff 1971; Sankoff and Laberge 1973, etc.).3 3 Within this general approach to creole genesis, two views have been advocated: the monogenetic and the polygenetic theory of creole languages. The former approach claims that European-based creoles are derived from a single pidgin, the Portuguese pidgin that emerged during the 15th century on the route of the Portuguese merchants. Among the tenants of this theory, we find Alleyne (1971), Goodman (1964), Hancock (1968), Stewart (1962), Whinnom (1956, 1965, 1971), to name but a few. On this view, the original pidgin would have been diversified with the dispersion of its speakers in the various countries of colonisation. As it became the native language of a first generation of speakers, this original pidgin language would have evolved into mutually unintelligible creoles due to borrowing from different

THE GENESIS OF PIDGIN AND CREOLE LANGUAGES

15

This theory accounts for a number of the characteristics listed in section 2.1. It accounts for the fact that pidgins emerge in multilingual communities (the first feature), for the fact that the members of communities where pidgins emerge are in need of a lingua franca (the second feature), and for the fact that creoles emerge rapidly, in this case in one generation (the fourth feature). However, it does not account for the fact that substratum speakers have little access to the superstratum language (the third feature), nor for the isolating character of pidgin and creole languages (the fifth feature), nor for the type of mix that pidgins and creoles manifest with respect to their source languages (the sixth feature). Finally, as has been pointed out in Lefebvre and Lumsden (1989: 253), without linguistic criteria distinguishing between pidgin and creole languages, the theory is not falsifiable (the seventh criterion). Precise definitions of pidgin and creole languages are desirable at this point. Pidgins and creoles have long been considered as separate entities on the basis of the following two sets of criteria. While pidgins have been defined as reduced codes, creoles have been defined as expanded versions of these reduced codes (see e.g. Hymes (ed.) 1971, and the references therein). Also, while pidgins have been found to often constitute the second language of the speakers who use them, a creole is considered to be a pidgin that has become the first language of a new generation of speakers, as we saw above (see also Kay and Sankoff 1974; Sankoff and Laberge 1973). In more recent literature, the distinction between pidgins and creoles has been levelled out in view of the fact that there are some pidgins (still used as a second language) that have been shown to have expanded in the same way as languages known as creoles (see e.g. Mühlhäusler 1980, 1986a, for an extensive discussion of this point). Hancock (1980: 64) states: "I prefer not to acknowledge a distinction between pidgin and creole, and to consider stabilisation more significant than nativisation in creole language formation." Similarly, Mufwene (1990: 2) uses the term creole to refer "to varieties traditionally called creoles but also to those called pidgins that serve as vernaculars or primary means of communication for at least a portion of their speakers." Moreover, in recent literature in the field, scholars have started referring to pidgins and creoles as PCs, suggesting that they fall into a single category. Furthermore, pidgin and creole languages cannot be distinguished on the basis of the processes that play a role in their formation (see

colonial languages (see e.g. Stewart 1967: 47). The polygenetic theory of creole genesis (see e.g. Hall 1966), stipulates that different pidgins gave rise to different creoles. In this view, a French-based pidgin would be the source of French-based creoles, an English-based pidgin would be the source of English-based creoles, and so on and so forth. The debate on this issue has raised enormous methodological problems. Some of these are discussed in Frake (1971), Goodman (1971), Grimshaw (1971), Hymes (1971), Southworth (1971), etc. To the best of my knowledge, no one advocates a monogenesis theory of PCs anymore.

16

ISSUES IN THE STUDY OF PIDGIN AND CREOLE LANGUAGES

Woolford 1983, for a general discussion of this point). Indeed, the processes hypothesised to play a role in the formation and development of human languages apply to both pidgins and creoles (see Keesing 1988; Lefebvre 1998a, and the references therein). Since these languages cannot be distinguished on the basis of these processes, no distinction should be made between them (see Lefebvre 1998a: 4). As will be seen further on, this will turn out to be a major drawback to Bickerton's (1981) Language Bioprogram Hypothesis which requires that pidgins and creoles be different and separate entities produced by different processes. 2.2.3 The theory of PCs as crystallised varieties of 'imperfect' second language acquisition According to the theory of imperfect second language acquisition of PC genesis (see e.g. Alleyne 1971, 1980; Andersen 1980, 1983; Chaudenson 1979, 1993; Mufwene 1990; Schumann 1978; Thomason and Kaufman 1991; Valdman 1980), PCs constitute the crystallisation of an imperfect version of the acquisition of a second language. In this view, the speakers of a hypothesised proto-creole lacking sufficient access to the colonial language data which they were exposed to would have created an approximate simplified system of the type of that found in some cases of second language acquisition.4 Does this theory account for the features of PCs enumerated in section 2.1? While this theory of creole genesis accounts in a straightforward way for the fact that speakers of the substratum languages do not have much access to the superstratum language in contexts where pidgin and creoles emerge (the third feature), it fails to account for several of the other characteristics enumerated in section 2.1. It does not account for the fact that PCs emerge only in multilingual communities (the first feature), nor for the fact that these communities need a lingua franca (the second feature), nor for the fact that PCs are created rather rapidly (the fourth feature), nor for the fact that PCs tend to be isolating languages (the fifth feature). More importantly, this theory does not provide an explanation for why PCs have 'crystallised' in the way they have with respect to their source languages (the sixth feature). Finally, this theory does not satisfy the seventh criterion, as it does not appear to be falsifiable. As Lefebvre and Lumsden (1989: 254) point out, this theory "ne définit pas précisément les mécanismes d'acquisition d'une langue seconde, ni en quoi elle diffère de l'acquisition d'une langue maternelle." [does not define in a precise way the mechanisms of second language acquisition, nor its difference with first

4

For a discussion of similar and contrastive properties of second language acquisition and creolisation, see e.g. Véronique (1994).

THE GENESIS OF PIDGIN AND CREOLE LANGUAGES

17

language acquisition]. Hence, it is not falsifiable (but see section 2.4 for refinements of this approach). 2.2.4 The theory of PCs as restructured varieties Several proposals may be regrouped under the view that PCs are restructured varieties. The following discussion is organised around three major proposals: PCs as restructured substratum varieties, PCs as restructured superstratum varieties, and PCs as restructured varieties of both of their source languages. 2.2.4.1 PCs as restructured substratum varieties. The substratist theory of creole genesis postulates that Caribbean creole languages have emerged by means of the gradual transformation of the West African languages (spoken by the slaves) influenced by the European colonial languages (see e.g. Alleyne 1980; Holm 1988). How does this theory meet the criteria in section 2.1? This theory may account for the fact that creoles only emerge in multilingual communities where there is a need for a lingua franca, and where speakers of the substratum languages have little access to the superstratum language (the first three characteristics in section 2.1). However, it does not provide an explanation for why creole languages are created in a relatively short period of time, nor for why they tend to be isolating languages (the fourth and fifth characteristics). While the postulated gradual transformation of the substratum languages influenced by the colonial languages does account for the contribution to the creole of both the substratum and the superstratum languages, it does not predict the principled respective contribution of these languages to the creole (the sixth characteristic). Finally, this theory is not falsifiable (the seventh characteristic), for, as has been pointed out in Lefebvre and Lumsden (1989: 254), it does not account for the facts that distinguish the emergence of PCs from cases of regular change occurring in languages (or in language contact varieties) that are not known as PCs. 2.2.4.2 PCs as restructured superstratum varieties. The superstratist theory of PC genesis holds that PCs constitute restructured dialects of their superstratum language. For example, in this view, French-based creoles would constitute restructured dialects of French, and so on and so forth. The main advocate of this approach is Chaudenson (1973, 1983, 1992). This approach may be said to account, to a certain point, for the fact that speakers of the substratum languages have little access to the superstratum language (the third feature). It does not account, however, for the fact that creoles only emerge in multilingual communities, where there is a need for a lingua franca (the first two features). It does not account for the relatively rapid

18

ISSUES IN THE STUDY OF PIDGIN AND CREOLE LANGUAGES

formation of PCs (the fourth feature) nor for the fact that PCs tend to be isolating languages (the fifth feature). Furthermore, and more importantly, it does not account for the principled division of properties of PC lexicons between their source languages (the sixth property). Is this theory falsifiable? This theory is falsifiable on the following grounds. As is extensively argued in Lefebvre (1998a, 2001d), PCs tend to reproduce the semantic and syntactic features of their substratum languages and hence, from a typological point of view, they pair with their substratum languages rather than with their superstratum languages. On this view, Atlantic creoles tend to reproduce the features of their West African substratum languages (see e.g. Lefebvre 1998a, and the references therein), whereas Pacific creoles tend to reproduce those of their Austronesian substratum languages (see e.g. Keesing 1988; Sankoff 1991). Thus, from a typological point of view, PCs resemble their substratum languages in spite of the fact that the phonological representation of their lexicons are derived from their respective superstratum languages. For example, as has been demonstrated in detail in Lefebvre (1998a), although the bulk of the phonological representations of Haitian words are derived from French, the typological features of Haitian pair with West African languages, not with French which shares features with Romance languages. This situation argues that Haitian Creole cannot be considered to be a dialect of French, for dialects of a given language are expected to share typological features. In my view, this situation falsifies the theory according to which PCs constitute dialects of their superstratum language. (For further discussion of this issue, see also Mufwene 1996b: 166.) 2.2.4.3 PCs as restructured varieties of both of their source languages. The idea that pidgins and creoles constitute restructured varieties of their substratum or superstratum languages has given rise to a recent collection of papers edited by Neumann-Holzschuh and Schneider (2000) under the title Degrees of restructuring in creole languages. This volume contains various papers presenting case studies of pidgins and creoles analysed as restructured varieties. Some papers propose that creoles are restructured varieties of both their substratum and their superstratum sources (see e.g. Alleyne 2000; Chaudenson 2000). According to some authors, creoles may vary with respect to degrees of restructuring (see e.g. Baker 2000; Holm 2000; Winford 2000). In their introduction to the collection, the editors point out the confusion regarding the definition of 'restructuring' and related concepts. With the exception of the fact that some authors now acknowledge the contribution of both substratum and superstratum sources to the creole, the framework adopted for the papers in the aforementioned collection presents the same problems as the

THE GENESIS OF PIDGIN AND CREOLE LANGUAGES

19

two positions discussed in sections 2.2.4.1 and 2.2.4.2. with respect to the features that any theory of PC genesis must be able to account for. 2.2.5 The theory that creoles reflect the properties of Universal Grammar The main proponent of the claim that creoles reflect the properties of Universal Grammar is Bickerton (1981, 1984, 1986, and subsequent work). (Note that, in this view, pidgins and creoles are crucially different entities, see below.) This theory, known as the Language Bioprogram Hypothesis (LBH), rests on the alleged similarity between undoubtedly historically unrelated creoles, such as Haitian, Sranan and Hawaiian, and on the alleged similarity between creoles and child language. According to Bickerton, each person is born with a grammatical model (the Language Bioprogram) enabling him or her to construct a grammar. In contexts where creoles emerge, children are exposed to a pidgin spoken by their parents. According to Bickerton, this pidgin is an impoverished language variety that does not present all the characteristics of a native language. Being faced with this impoverished linguistic model, the children use their hypothesised Language Bioprogram in order to nativise the pidgin. Nativisation of the pidgin is claimed to consist in expanding the pidgin. The language variety so created is claimed to be a creole that reflects both the unmarked grammar that is hypothesised to characterise the language of young children, and the unmarked grammar that is hypothesised to characterise creole languages. Thus, in Bickerton's view, both creole languages and child language are closer to Universal Grammar than other language varieties, for both present the unmarked options of Universal Grammar (see also Seuren and Wekker 1986 for a similar view on this point). Still on this view, in ordinary cases of first language acquisition, children are exposed to linguistic data that are produced by the adults around them. Presumably, in this situation, children have a chance of acquiring the language specific features of their native language. Bickerton claims that, in the special case of first language acquisition in the context of creole genesis, children are deprived of an adequate adult model, in such a way that the language that they develop has the features of Universal Grammar. Bickerton claims that his theory accounts for both the hypothesised similarity between creole languages and the hypothesised similarity between creoles and child language.5 Does this theory account for the seven properties identified in section 2.1? Since Bickerton crucially considers pidgins and creoles as separate entities, 5 Note that Bickerton's (1981) Language Bioprogram and Chomsky's (1986) Language Acquisition Device are sometimes considered as being equivalent. In my understanding of the two approaches, however, these two devices are quite different. On Bickerton's view, it is the lack of pertinent data that activates the Language Bioprogram, whereas on Chomsky's view, it is the presence of pertinent data that activates the Language Acquisition Device. It thus seems that the two devices are not equivalent with respect to the trigger that activates them. The association of these two devices thus appears to be erroneous.

20

ISSUES IN THE STUDY OF PIDGIN AND CREOLE LANGUAGES

I will refer only to creoles in addressing this question. Bickerton's approach does not account for the fact that creole languages emerge only in multilingual communities that are in need of a lingua franca and where language learners have little access to the superstratum language (the first three features). Since the creole is nativised in one generation, this theory can be said to account for the rapid development of creole languages (the fourth feature). The theory does not account, however, for the fact that pidgin and creole languages tend to be isolating languages nor for the fact that they manifest the properties of both their substratum and superstratum languages in the way they do (the fifth and sixth features). The theory is formulated in terms that are precise enough so as to be falsifiable. Building on Lefebvre and Lumsden (1989: 255), the next paragraphs discuss several points that falsify the universalist approach to creole genesis. Crucially, the Language Bioprogram Hypothesis of creole genesis requires that pidgins and creoles be different entities formed by different processes. On the one hand, it has been demonstrated that pidgins and creoles are not qualitatively different from one another (see in particular the work of Sankoff, e.g. Sankoff and Brown 1980; Sankoff and Laberge 1980). On the other hand, as was mentioned in section 2.2.2. and, as will be shown in section 2.4, pidgins and creoles are not distinguishable on the basis of the processes that are at work in their formation. This constitutes a first major drawback to Bickerton's theory. Second, the alleged similarity between creole languages falls short in view of detailed comparisons of various creoles. As is shown in Muysken (1988b), while a superficial look at creole languages may yield the conclusion that they are alike, a closer look at the data forces a revision of this conclusion. An example in point is the fact that, while some creoles manifest the serial verb construction (e.g. Saramaccan, Jamaican, Haitian, Papiamento, Tok Pisin, etc.), others do not (e.g. Philippine Creole Spanish, Hawaiian Creole English, Mauritian Creole, Seychellois, Reunionais, etc.). Likewise, while some creoles manifest the predicate cleft construction (e.g. Haitian, Papiamento, etc.), others do not (e.g. Tok Pisin, Solomons Pidgin, Australian creoles, etc.). Furthermore, as is extensively discussed in Lefebvre (1998a, 2001d), creoles tend to reproduce the semantic and syntactic features of their substratum languages. Hence, Atlantic creoles tend to reproduce the features of their West African substratum languages, whereas Pacific creoles tend to reproduce the features of their Austronesian substratum languages. This explains why Atlantic creoles manifest the predicate cleft construction, whereas Pacific ones do not. A comparison of Haitian (and contributing languages) in Lefebvre (1998a) with Solomons Pidgin (and contributing languages) in Keesing (1988) strongly supports this claim. In conclusion, the alleged similarity between creole

THE GENESIS OF PIDGIN AND CREOLE LANGUAGES

21

languages falls short when data from creoles of different geographical areas are considered. Third, the hypothesised similarity between creole languages and child language receives no support in current literature (see for example the critiques formulated by a number of scholars in a special issue of Brain and Behavioral Sciences, 1984). Fourth, the claim that creole languages reflect the unmarked case is not of much use without a theory of markedness. Indeed, no theory of markedness is formulated in Bickerton's work. Furthermore, based on a theory of markedness, some authors show that creole languages do present marked options of Universal Grammar. Indepth discussions of this point can be found in Koopman (1986), Lefebvre (1998a) and Muysken (1981b). Fifth, Bickerton's theory loses even more points when historical data are considered. For example, Singler (1996) shows that nativisation of the Caribbean plantation societies was an extremely slow process. First, the slave traders imported twice as many men than women (Curtin 1976); second, the birth rate was very low (Kiple 1984); third, infant mortality was very high (Singler 1993a); fourth, life span of Africans in the Caribbean was short (Singler 1993a). As Singler (1993a: 237-238) comments: "This combination of factors yielded societies unable to reverse the natural population decrease. They were societies marked by both a disproportionately small number of children and an ongoing stream of recently arrived slaves from Africa". Since the bulk of the Caribbean population at the time the creoles were formed was adult, Singler (1996: 199) concludes that the principal agents of creole genesis must have been adults. Moreover, as will be shown in section 2.4, the main process at work in the formation of pidgin and creole languages requires adult language competence. In conclusion, the Language Bioprogram Hypothesis of creole genesis does not hold in view of all these facts. 2.2.6 Summary In this section, six approches to creole genesis were reviewed. All share the characteristic that they focus on language varieties rather than on processes that lead to these language varieties. Each of these theories has been evaluated against the seven features that need to be accounted for by any theory that seeks to provided a complete theory of the origin of these languages, as per the criteria established in section 2.1. The results are summarised in Table 2.1, which should be interpreted in light of the methodological proviso at the end of the introduction to section 2.2.

22

ISSUES IN THE STUDY OF PIDGIN AND CREOLE LANGUAGES

Table 2.1. Summary of the six theories of pidgin and creole genesis with respect to the features that need to be accounted for by any theory that seeks to account for the origin of these languages. (The symbols + and - indicate whether a given theory accounts for a given feature) Features→

theories↓ PCs as reduced codes Creoles as nativised pidgins PCs as cases of imperfect second language acquisition PCs as restructured substratum varieties PCs as restructured superstratu m varieties LBH

Multi lingual com munity

Need for Little access a lingua to superfranca stratum language

Rapid creation

Isolating language

Type of mix

Theory is falsifiable

-

-

+

-

+

-

+

+

+

-

+

-

-

-

-

-

+

-

-

-

-

+

+

+

-

-

-

-

-

-

+

-

-

-

+

_

_

_

+

_

_

+

As can be seen from Table 2.1, none of the six theories reviewed in this section can account for all the features that need to be accounted for in an optimal theory of PC genesis. Strikingly enough, while each theory accounts for at least one of the features in the list, none of the theories evaluated thus far can account for the type of mix that is manifested by PCs from among their source languages. As will be seen in section 2.4, the relexification account of PC genesis crucially predicts the principled division of properties of PC lexicons between their source languages. With the exception of the nativised pidgin theory of creole genesis (section 2.2), all the theories discussed in this section consider PCs as somewhat deprived language varieties (see e.g. the 'baby-talk' or the 'foreigner-talk' approach, the 'imperfect' stage of second language acquisition, the 'restructured' varieties, the alleged similarity between creole languages and child language). For a long time, PCs were considered 'marginal'. To my knowledge, the first state-of-the-art article on PCs was written as late as 1964; it was entitled 'Trade

THE GENESIS OF PIDGIN AND CREOLE LANGUAGES

23

Jargons and Creole Dialects as Marginal Languages', signed by Reineke.6 In the early seventies, several linguists such as Hall (1966: 121-122), Labov (1971), Whinnom (1971: 109), etc., urged students of PCs to approach these languages in the same way as they do approach other languages. The first substantial collection of papers on pidgins and creoles was edited by Dell Hymes in 1971. This collection is still an extremely valuable source. In the late seventies and early eighties, we begin to see discussions of pidgin and creole genesis within the framework of the processes otherwise known to play a role in language genesis and language change in general. 2.3.

Shifting the object of study from language varieties to processes involved in language creation and change In this section, I undertake the discussion of what I assume to be the second phase of research on pidgin and creole genesis. I estimate this second phase to have begun in the late seventies. It is characterised by the desire to cast the discussion of PC genesis within the framework of the processes otherwise known to be at work in language formation and in language change in general. On the one hand, proposals on the origin of languages referred to as mixed languages7 are brought into the forum of discussion on the origin of PCs. In this respect, Media Lengua—a Quechua-Spanish mixed language spoken in Ecuador—(see Muysken 1981a, 1988c), Michif—a Cree-French mixed language—spoken by the metis buffalo hunters of Canada and the Northern United States (see Bakker 1989, 1992, 1994; Papen 1988), and Inner Mbugu or Ma'a—a mixed language spoken in Tanzania (see Goodman 1971; Möhlig 1983; Mous 1994, 1995, in press; Thomason and Kaufman 1991), to name but a few, and the theories of their emergence, were brought to the scene of PC studies. On the other hand, scholars started to entertain the idea that the changes observed in pidgin and creole languages are not fundamentally different from regular cases of language change in non-creole languages (see e.g. Adone and Plag (eds) 1994; Baker and Syea (eds) 1996; Hymes (ed.) 1971; Plag 1994a; Sankoff (ed.) 1980; etc.). Finally, sociolinguistic studies seeking to explain the formation of urban varieties out of various rural ones (see e.g. Domingue 1980, 1981; Jourdan 1985; Siegel 1995, 2002; Trudgill 1986) also had their input into PC studies. Three major processes have gradually entered the scene of PC studies. These processes are relexification, shown to play a role in the formation of mixed 6

Thanks to John Reineke for having brought PCs to the attention of linguists. Languages that are being referred to as mixed languages emerge in contexts where only two languages are spoken in contrast to PCs which emerge in contexts involving more than two languages, as we saw in section 2.1. For an extensive discussion of the differences and similarities between PCs and mixed languages, and the situations in which they emerge, see Lefebvre (1998a: 29-30, and the references therein). 7

24

ISSUES IN THE STUDY OF PIDGIN AND CREOLE LANGUAGES

languages; reanalysis, a major process in linguistic change; and dialect levelling, a process that takes place when various dialects of the same language come into contact. In this section, I provide definitions for these processes and I show how they began to be applied to PC formation and development. The shift from the study of language varieties to the study of the processes at work in the formation of these varieties constitutes a positive step in the study of PC genesis. As will be seen in the following subsections, however, this shift was not entirely successful on the first round. 2.3.1 Relexification Muysken (1981a, 1988c) shows that Media Lengua has a lexicon where the phonological forms of major syntactic category lexical items (that is, nouns, verbs and adjectives) are almost entirely derived from Spanish, while the forms of the affixes and of the functional category system are derived from Quechua. The examples in (2) and (3) illustrate this division. (In the Media Lengua examples below, the forms derived from Spanish appear in italics, and those derived from Quechua, in regular characters.) (2)

a.

No

sé.

not know. 1st T do not know.' b. Mana yacha-ni-chu. No sabi-ni-chu. not know. 1st.VAL T do not know.' (3)

SPANISH

QUECHUA MEDIA LENGUA (=(3) in Muysken 1981a)

a. Si llueve demás, no voy a ir. SPANISH if rain.3rd too.much, not go.1st to go Tf it rains too much, I will not go.' b. Yalli-da tamia-pi-ga, mana ri-sha-chu. QUECHUA Dimas-ta llubi-pi-ga, no i-sha-chu. MEDIA LENGUA too.much-ACC rain-LO-TO not go-ASP-VAL Tf it rains too much, I will not go.' (=(1) in Muysken 1981a)

Muysken (1981a) demonstrates that although the major category lexical entries of Media Lengua derive their phonological representations from Spanish, their semantic content is derived from Quechua. For example, the phonetic form of the Spanish verb sentarse 'sit down' was used to replace the Quechua verb tiyari 'sit', 'live', 'locative be', 'there is' yielding the Media Lengua verb sinta-ri 'sit', 'live', 'locative be', 'there is'. As Muysken (1981a: 56) points out, a single Media Lengua word is substituted for the Quechua word, preserving the various meanings of the latter even when, in Spanish, each of these meanings would be expressed by a separate lexical item: estar sentado 'sit', vivir 'live',

THE GENESIS OF PIDGIN AND CREOLE LANGUAGES

25

estar 'locative be', hay 'there is'. Similarly, the phonetic form of the Spanish verb tener 'to have, to hold' was used to replace the Quechua verb chari- 'to have' yielding the Media Lengua verb tini- 'to have' (see Muysken 1988c). The above examples show that, although there is some overlap in the semantics of the verbs that are associated in the replacement, the semantics of the new Media Lengua verbs follows the details of Quechua rather than those of the Spanish lexical entries. What is the process that produces lexical entries having the division of properties of the type described above? Muysken (1981a: 61) identifies this process as relexification: "Given the concept of lexical entry, relexification can be defined as the process of vocabulary substitution in which the only information adopted from the target language in the lexical entry is the phonological representation." Muysken's representation of the process is reproduced in (4).

In this view, relexification is a mental process that builds new lexical entries by copying the lexical entries of an already established lexicon and replacing their phonological representations with representations derived from another language. The nature of this process accounts for the division of properties observed in Media Lengua. The process of relexification has been argued to be an important tool in the creation of other mixed languages (e.g. for Michif, see Bakker 1989, 1992, 1994; for Inner Mbugu or Ma'a, see Mous 1994, 1995, in press).8 This process has also been claimed to play a role in PC genesis. For example, Koopman (1986), Lefebvre (1984, 1986), Stewart (1962), Voorhoeve 8

For a summary of the role of relexification in the formation of various mixed languages, see Lefebvre (1998a: 18-29).

26

ISSUES IN THE STUDY OF PIDGIN AND CREOLE LANGUAGES

(1973), Whinnom (1977) and others have long claimed that this is so. At one point, Muysken (1981a: 77) also proposed that relexification played a role in the formation of PCs: "If it is the case that the Caribbean creoles show numerous African survivals in their syntax and semantics, then I think we can argue that it is not interference which led to these survivals, but relexification." In a more recent paper (Muysken and Smith 1990: 884), however, Muysken amends his earlier claim, allowing relexification to play a role in language genesis only in bilingual situations, that is, only in the formation of mixed languages. "We reject the gradual 'relexification' of believers in monogenesis (from a West African Portuguese Pidgin) or Afrogenesis, in situations of communal linguistic confrontation between, e.g. a European planter class and an African slave class. We do accept the possibility of relexification as a mechanism in forming a new language in a bilingual situation." This strong position had the effect of slowing down research on the role of relexification in PC genesis for a while, but, as will be seen in section 2.4, it did not have the effect of killing the idea altogether. 2.3.2 Reanalysis and related phenomena Reanalysis (and related phenomena, e.g. desemanticisation and grammaticalisation)9 constitutes a major process of linguistic change (see e.g. Heine and Reh 1984; Hopper and Traugott 1993; Lightfoot 1979, etc.). Reanalysis is a mental process by which a particular form which signals a lexical entry becomes the signal of another lexical entry (see e.g. Lightfoot 1979). 10 A typical example is the reanalysis of verbal expressions as adverbs. For example, the Yoruba verbal expression sa ere 'run race' has been reanalysed as an adverb: sere 'quickly' (see Bámgbósé 1974; Lord 1976). Since the seventies, several cases of linguistic change that have occurred in PCs have been analysed as cases of reanalysis. Such cases are reported in Baker and Syea (eds) (1996), Bickerton (1988), Foley (1988), Koopman and Lefebvre (1981), Lefebvre (1984), Mühlhäusler (1986a), Muysken (1988b), Plag (1993), Rickford (1987), Romaine (1988), Sankoff (1990, 1991), Sankoff

9 There is an ongoing debate on whether cases of grammaticalisation and of reanalysis constitute a single process or two separate ones. For example, Heine and Reh (1984: 97), Hopper and Traugott (1993: 32) and Lefebvre (1998a: 41-45) consider that reanalysis subsumes cases of grammaticalisation. Haspelmath (1998, 1999), however, strongly argues for formal differences between them. Since nothing in the present discussion hinges on possible distinctions between reanalysis and grammaticalisation, I do not pursue the discussion of this point here. 10 I take the process of reanalysis to apply within a particular language. There are other acceptions of the term, however. For example, some authors consider calques as cases of reanalysis. Such cases are not included in my use of the term reanalysis.

THE GENESIS OF PIDGIN AND CREOLE LANGUAGES

27

and Laberge (1973), Valdman and Highfield (eds) (1980), Washabaugh (1975), to name but a few.11 However, most of these authors were generally looking at data drawn from the pidgin or creole alone, and data from contributing languages to these PCs were generally not considered. As a consequence of this, more cases of reanalysis were postulated to have occurred in PCs than had actually taken place. Indeed, as was subsequently shown by various authors, when the properties of the corresponding lexical entries in the substratum languages are considered as a point of departure for the pidgin or creole lexical entries, there is less need to call upon reanalysis than was first hypothesised to explain the various functions of a particular lexical item in PCs (see e.g. Bruyn 1996; Keesing 1988; Lefebvre 1998a; Traugott 1999). For example, Koopman and Lefebvre (1981, 1982) hypothesised that the complementiser pou entered the Haitian lexicon through a process of reanalysis involving both the preposition pou 'for' and the irrealis mood marker pou. Lefebvre (1998a: 191-193) shows, however, that the corresponding substratum lexical entries cumulate these functions. It thus appears that, in this particular case, there is no need to have recourse to reanalysis to explain the creole data, and that relexification alone turns out to account for the full range of functions of the creole lexical entry. 2.3.3 Dialect levelling Dialect levelling, as discussed in the literature on dialects in contact (see e.g. Domingue 1980, 1981; Siegel 1985, 1987, 1995, 1997; Trudgill 1986; etc.) refers to the reduction of variation between dialects of the same language in situations where these dialects are brought together. As Siegel (1997: 21) puts it, "dialect differences are reduced as speakers acquire features from other varieties as well as avoid features from their own variety that are somehow different. This may occur over several generations until a stable compromise dialect develops." Well-documented cases of dialect levelling include Bhojpuri as spoken in Mauritius (see Domingue 1980, 1981) and English as spoken outside of England (see e.g. Siegel 1997; Trudgill 1986). In the recent literature on pidgins and creoles (mainly in the eighties and nineties), it has been suggested that dialect levelling also plays a role in the further development of these languages (see e.g. Harris 1991: 199; Jourdan 1985; Mufwene 1990: 138-139, 1994c, 1996b: 22; Mühlhäusler 1980: 34; etc.). For example, Siegel (1997: 26) asserts that: "Mixing and levelling may (...) be important in the development of stable pidgin and creole languages... When the various versions of the superstrate are then used as the main means of 11 For an extensive discussion of how cases of reanalysis reported to have taken place in PCs are of the same kind as those observed in other languages, see Lefebvre 1998a: 30-33).

28

ISSUES IN THE STUDY OF PIDGIN AND CREOLE LANGUAGES

communication among speakers of different substrate languages (in other words, when vernacularisation occurs), and when these speakers begin to view themselves somehow as a 'community', then levelling begins." With the exception of Siegel who claims that variation within a given PC comes from the various second language acquisition versions of the lexifier language, at this stage, no precisions are given as to where the variation found in PCs comes from. Documented cases of dialect levelling in PC development are quasi non-existent. 2.3.4 Summary In the late nineteen seventies and in the eighties there was a shift in the focus of enquiry about PC genesis from the study of language varieties to the study of the processes yielding these varieties. Three major processes shown to play a role in language genesis and language change in general—relexification, reanalysis and dialect levelling—were hypothesised to also play a role in the creation and development of PCs. As we saw earlier, the use of these processes in the study of PC genesis was not entirely successful on the first round. There are two major reasons for this situation. The first one is due to the lack of systematic and detailed comparative work between PCs and their source languages. For example, the overestimation of the role of reanalysis mentioned in section 2.3.2 is without doubt due to this factor. The second reason is due to the fact that there was no theory on how these three processes apply and interact in PC genesis. Each process was being considered for its own sake. Since none of the three processes taken individually may constitute a theory of PC genesis12, at the end of this second phase, the field was left, so to speak, in a state of flux. The turning point into the third phase of studies on PC genesis was marked by the debate between substratists and universalists (see Muysken and Smith (eds) 1986) and by Keesing's (1988) publication of the first comparison of a PC with its source languages: the comparison of Solomons Pidgin with English, its lexifier language, and with Kwaio, one of its Austronesian substratum languages. At the same time, another piece of comparative research had been undertaken on an Atlantic creole, Haitian Creole, to which I now turn. 2 . 4 . A unified theory of pidgin and creole genesis From 1985 to 2000, successive projects on the genesis of PCs were carried out at the Université du Québec à Montréal (hereafter UQAM).13 These projects were based on the general assumption that it should be possible to account for the formation of PCs in terms of the processes that are at work in language 12 For this reason it is not possible to discuss the proposals reviewed in this section against the seven features in section 2.1. 13 A summary of the history of these projects can be found in the Preface to Lefebvre (1998a).

THE GENESIS OF PIDGIN AND CREOLE LANGUAGES

29

genesis and language change in general, that is relexification, reanalysis and dialect levelling, and in terms of a sound theory of how these processes interact in PC genesis (see Lefebvre 1986, 1993b, 1998a, and the references therein; Lefebvre and Kaye (eds) 1986; Lefebvre and Lumsden 1989, 1994a, 1994b). Another assumption was that the linguistic account should be compatible with the situation that prevailed at the time the creole was formed. Because of the importance of this research in terms of resources14 and production15, this section is dedicated to an overview of the theoretical progress in the account of the genesis of PCs accomplished within the framework of the aforementioned projects. Section 2.4.1 presents an overview of the hypothesis and of the methodology of this research. Section 2.4.2 summarises the contribution of this research to the issue of how the processes involved apply and interact in PC genesis. Section 2.4.3 evaluates the proposed theory on the basis of the characteristics that define an optimal account of creole genesis as provided in section 2.1. 2.4.1 Hypothesis and methodology of the UQAM projects The basic hypothesis (as formulated in Lefebvre 1986, 1993b; Lefebvre and Kaye (eds) 1986; Lefebvre and Lumsden 1989, 1994a, 1994b; etc.) tested by the research reported on in this section is that the creators of a creole language, adult native speakers of various languages, use the properties of their native lexicons, the parametric values and semantic interpretation rules of their native grammars in creating a PC. The bulk of a PC's lexical entries is created by the process of relexification. Two other processes, fed by the output of relexification, dialect levelling and reanalysis, also play a role in the development of a PC. As is pointed out in Lefebvre and Lumsden (1994a), this account is a further development of the second language acquisition theory of PC genesis (see section 2.2.3): it is claimed that, in creole genesis involving situations where there is little access to the superstratum language, the process of relexification is used by speakers of the substratum languages as the main tool for acquiring a second language: the superstratum language. The hypothesis was tested on the basis of Haitian Creole. The research programme involved two dimensions, a historical dimension and a linguistic one. The historical research was designed to answer the following questions: when was Haitian Creole formed? What were the salient demographic characteristics of the Haitian population during that period? Who were the people present at the relevant time? What was their linguistic background? (see Lefebvre 14 For a complete statement of the human and financial resources allocated to these projects, see the Preface to Lefebvre (1998a). 15 See Lefebvre (1998a) and Lefebvre and Brousseau (2002, and the references therein).

30

ISSUES IN THE STUDY OF PIDGIN AND CREOLE LANGUAGES

1993b). The historical research in the colonial archives of France was carried out by John Singler (see Singler 1993a, 1993b, 1996). In short, Singler establishes the following points. Haitian Creole was formed between 1680 and 1740. As a consequence of a shift from a tobacco and cotton economy to a sugar economy, the number of colonists decreased and the number of slaves exploded; this had the effect of modifying the slave population's exposure to French. The bulk of the Caribbean population at the time Haitian Creole was formed was adult (see details in section 2.2.5). As for the languages that these adults were speaking, Singler (1993b) shows that they were all Niger-Congo languages, more particularly Kwa (Gbe and Akan) and Bantu. During the formative period of Haitian Creole, Gbe speakers made up more than 50% of the French Caribbean slave-export population. As is pointed out in Lefebvre and Lumsden (1994b), the overall situation found in Haiti between 1680 and 1740 presented all the prerequisites for the emergence of a creole language: there was a multilingual community, in need of a lingua franca, and the bulk of the population, the speakers of the substratum languages, had only reduced access to the superstratum language. The test of the linguistic hypothesis consists in a detailed comparison of the lexicon and grammar of Haitian Creole with those of its contributing languages: French, its superstratum language, and West African languages, its substratum languages. Due to time and resource constraints, we decided to limit the detailed study of the substratum languages of Haitian to one language. Because of the importance of the influence of the Fon culture on that of Haiti (with respect to religion and art, see e.g. Bastide 1967; Herskovits 1975), Fongbe, a language of the Gbe cluster, was chosen as the substratum language to be studied in detail (see Lefebvre 1986, 1993b; Lefebvre and Kaye (eds) 1986).16 This choice turned out to be a good one, in view of Singler's finding that the Gbe speakers outnumbered speakers of the other West African languages at the time Haitian Creole was formed. As has been pointed out in Lefebvre and Lumsden (1994a), the methodological choices that were made had the effect of making the relexification hypothesis easier to falsify. (For a thorough discussion of the methodology of the research and the validity of the linguistic test, see Lefebvre 1998a: 52-77, and the references cited therein.) The linguistic test involves a global comparison of the lexicons, parametric values, semantic interpretation rules and concatenation principles of the languages involved. As has been emphasised in Lefebvre and Lumsden (1994a, 1994b), one or two examples either way are not enough to support or falsify the hypothesis. The test must involve quantity as well as quality. 16 In no way does this methodological choice entail that Haitian is Fongbe relexified. For further discussion of this point, see Lefebvre 1998a: 52-77.

THE GENESIS OF PIDGIN AND CREOLE LANGUAGES

31

To the best of my knowledge, this was the first time that sufficient resources have been gathered to make such a detailed and extensive comparison of the grammar and lexicon of a creole language with those of its superstratum and substratum sources. The global results of this threeway comparison can be found in Lefebvre (1998a). Additional results are reported on in Brousseau (in preparation), Lefebvre (1999a, 2001a), and in Lumsden (1999a, 1999b). The bulk of the threeway comparison supports the hypothesis in a way that surpasses my expectations when I started this research. 2.4.2 The interplay of the processes embedded within a scenario of creole genesis The content of this section summarises the theory developed during the 1989-1994 UQAM project with respect to how the three processes discussed in section 2.3 interact in a scenario of creole genesis. Unless otherwise specified, the theory reported on here is as developed in Lefebvre and Lumsden (1989, 1992, 1994a, 1994b) and in Lumsden and Lefebvre (1994). Relexification [also referred to as relabelling] applies in creole genesis in the following way. Native speakers of various substratum languages are brought together. Crucially, the speakers of the substratum community do not have a common language, a situation which creates the need for a lingua franca. The substratum speakers are exposed to a superstratum language, the language of the colonists. However, they do not have enough exposure to this language to learn the details of its lexical entries. Due to this situation, speakers of the substratum languages relexify the lexical entries of their respective lexicons on the basis of phonetic strings found in the superstratum language (see the examples in (1)). The relexification of various lexicons on the basis of a single superstratum language provides the speakers of the substratum languages with a common vocabulary. It is the limited direct access to the superstratum language that makes relexification so important in the formation of radical17 creoles. It is a well documented fact that in PCs, functional category items as well as major category lexical entries have phonological representations that are similar to some phonetic strings of the superstratum language. These lexical entries, however, do not have the same properties as the corresponding superstratum forms from which they are phonologically derived (see e.g. Carden and Stewart 1988; Lefebvre 1984; Lefebvre and Lumsden 1989, 1992; Mufwene 1991). In the scenario of creole genesis developed during the 1989-1994 UQAM project, it is hypothesised that, because speakers of the substratum languages have very limited access to the superstratum data, they typically fail to identify the functional categories of the superstratum language, 17

Creoles which less resemble their superstratum languages are referred to as radical creoles.

32

ISSUES IN THE STUDY OF PIDGIN AND CREOLE LANGUAGES

These speakers thus try to relexify the functional items of their native languages on the basis of forms found in the superstratum language. It is proposed that the functional category lexical entries of the substratum languages are relexified on the basis of major category lexical items (e.g. nouns, adjectives, verbs, adverbs and prepositions) of the superstratum language. For example, the definite determiner of the substratum languages of Haitian Creole is argued to have been relexified on the basis of the postposed French adverb là, yielding Haitian la (see Lefebvre 1998a: 79-84). The relexification of functional, as well as major category lexical entries, provides the speakers of the various substratum languages with a common vocabulary in all areas of the lexicon. The process of relexification is semantically driven in the sense that there must be partial semantic overlap between the source and target lexical entries for it to take place (see Muysken 1981a). Consequently, relexification is constrained by what the superstratum language has to offer in terms of appropriate strings to relexify original lexical entries. This is particularly crucial in the case of functional category items. It is thus possible that some lexical entries cannot be assigned a new phonological form in relexification, either because an original lexical entry has no semantic content (e.g. operators, case markers) or because there is no form available in the superstratum language to provide a new phonological form for an original lexical entry. Due to space limitations, such cases are not discussed here. I refer the reader to Lefebvre (1998a) for extensive discussion of such cases. The lexicons created by relexification become the basis of a lingua franca within the creole community. When the relexified lexicons become the target of the creole community, a new language is born. At this point, the speakers are no longer targeting the superstratum language. They are targeting the common language that they have developed through relexification: the incipient creole. At this stage, two other processes come into play: dialect levelling and reanalysis. Relexification is a cognitive hence individual process. Situations where creoles emerge involve several substratum languages. Each individual relexifies his or her own lexicon. Hence, speakers of various substratum languages reproduce the idiosyncratic semantic and syntactic properties of their own lexicons in relexification and thus, the product of relexification is not necessarily uniform across the creole community. The relexification of several lexicons thus creates variation within a creole. This scenario allows for a sound explanation of the facts referred to in the literature on creole studies as the 'cafeteria principle'—a term used first by Dillard (1970) and later by Bickerton. As Bickerton (1981: 49) puts it: "As things stand, we are asked to believe that different African languages contributed different rules and features to particular creoles (...) it is (...) absurd to suppose that a creole could mix fragments of

THE GENESIS OF PIDGIN AND CREOLE LANGUAGES

33

Yoruba, Akan, Igbo, Mandinka, and Wolof (...)." The differences created by the relexification of various lexicons may (but need not) be levelled out with time. The proposal that dialect levelling operates on the output of the various relexified lexicons involved in creole formation provides a principled explanation of the observation that several different substratum languages may contribute features to a given creole. Indeed, in the competition among different creole dialects (created by the relexification of different substratum lexicons), there are winners and losers. As is shown in Lefebvre (1998a), the competition is not always won by speakers of the same relexified lexicon (see also Siegel 1997). An original lexical entry that was not relexified (either because the creators of the creole did not find an appropriate form in the superstratum language to relexify the copied lexical entry, or because it had no semantic content and thus could not be relexified) may be signalled by a periphrastic expression. For example, a tense or aspect may be signalled by an adverb with a similar meaning. The periphrastic expression may later become the phonological representation of the functional category in question through the process of reanalysis. The postulated link between relexification and reanalysis accounts in a straightforward way for the paradoxical situation noted in the literature according to which, in the course of their further development, creoles develop lexical entries that manifest the properties of their substratum languages even in situations where the substratum languages have ceased to be spoken (see e.g. Chaudenson 1994; Mufwene 1990; Mühlhäusler 1986a, 1986b; Sankoff 1991: 73). (For an extensive discussion of this point, see Lefebvre 1998a: 108-110, 375-386.) 2.4.3 An optimal account of creole genesis The theory of creole genesis outlined above provides a straightforward and optimal account of all the properties of creole languages discussed in section 2.1. The following discussion builds on a preliminary one in Lefebvre and Lumsden (1989, 1994a), as updated in Lefebvre (1998a). The theory accounts for the fact that creole languages emerge in multilingual societies in need a lingua franca (the first and second features). It accounts for the fact that the creators of the creole have little access to the superstratum language (the third feature). Only one generation of speakers is required to create a new language by means of relexification (the fourth feature). When it is created, this new language evolves as any other language. By virtue of the definition of the process, creole lexical entries are predicted to have the same semantic and syntactic properties as the corresponding lexical entries in the substratum languages, but phonological representations derived from the phonetic strings of the superstratum language. The relexification theory of creole genesis thus accounts for the fact that creoles

34

ISSUES IN THE STUDY OF PIDGIN AND CREOLE LANGUAGES

reflect the properties of both their superstratum and their substratum source languages in the way they do (the sixth feature). As is observed in Lefebvre and Lumsden (1994a), the fact that creoles are generally isolating languages also follows from the above proposal. Since the functional category lexemes of creole languages derive their phonological forms from major-category lexemes in the superstratum language, or from reanalysis, and since these categories are typically free morphemes, it follows that creoles will tend to be isolating languages (the fifth feature). Finally, the theory that the bulk of a Creole's lexical entries are formed by the process of relexification is falsifiable. As is stated in Lefebvre and Lumsden (1989, 1994a), if the three-way comparison of the lexical properties of a radical creole with the lexical properties of its source languages were to show that the syntax and semantics of the creole are not systematically parallel to the syntax and semantics of the substratum languages, then the theory would be falsified. Likewise, the theory that the creators of the creole use the parametric values, the semantic interpretation rules and the principles of concatenation of their own grammars in creating the creole is falsifiable. If a comparison of the grammatical and semantic properties of a creole with those of its source languages were to show that the properties of the creole are not systematically parallel to those of the substratum languages, then the hypothesis would be falsified. Thus, the theory of creole genesis summarised in this section does account for all the features of an optimal theory of creole genesis. Consequently, if we were to add the above account of PC genesis to the list of theories in Table 2.1, all the features that need to be accounted for would be assigned a positive value. 2.4.4 Summary The core of the results of the UQAM Haitian projects can be found in Lefebvre (1998a, and the references cited therein). The detailed comparison of Haitian and its source languages overwelmingly supports the relexification account of creole genesis (see also chapter 3). The detailed comparison of Solomons Pidgin with its source languages by Keesing (1988) also supports such an account. The results in Migge (1998b) go in the same direction as well. As more cases are being documented, we deepen our understanding of the process itself and of the constraints upon it. 2.5. Conclusion Three major phases of investigation on PC genesis have been summarised in this paper. The first phase includes six major approaches. These approaches were shown to have in common the fact that they all focus on linguistic varieties. It was shown that none of them can account for all the features that characterise an

THE GENESIS OF PIDGIN AND CREOLE LANGUAGES

35

optimal theory of PC genesis. The second phase was characterised by an attempt to shift the focus of study from the language varieties to the processes that are at work in creating these language varieties. These processes are relexification, reanalysis and dialect levelling. For various reasons discussed in section 2.3, the first attempt at applying these processes to pidgin and creole genesis and development was not entirely successful. The third phase of studies in pidgin and creole genesis made a breakthrough in several respects. It provided a unified theory of the origin of PCs cast within the framework of the processes which otherwise play a role in language genesis and language change in general, and it provided a theory of how these processes interact in PC genesis. It also provided detailed and systematic comparisons of pidgins and creoles with their source languages.18 What is ahead in the study of pidgin and creole genesis? Surely, the documentation of more cases will provide new questions for the theory (see for example the papers in Siegel 2000). Types of studies that are needed for getting new insights into the theory are discussed in Lefebvre (2002), of which section 3 is here reproduced as Appendix 1. Regardless of what theory of PC genesis 18 At the symposium "Pidgin and Creole Linguistics in the 21st century" (1998), Glenn Gilbert asked participants the following question: "What place will universalist theories retain in 21st-century pidgin and creole linguistics?". My reply to his question was the following (see Lefebvre 2002: 247-286):

Universals of language should be assumed to be part of PCs on two grounds: first, by definition, language universals are universal; second, pidgin and creole languages are natural languages and therefore they reflect the universal properties of language just as other natural languages do. The differences betweenPCSand their contributing languages, on the one hand, and among PCS, on the other hand, are to be found in areas of the grammar that allow for variation between languages. The lexicon is the component par excellence where differences between languages exist. This has two major consequences. First, on the relexification account of creole genesis, creole lexicons, above all, should reproduce the idiosyncrasies of their substratum language lexicons; this prediction is borne out by the Solomons Pidgin discussed by Keesing (1988) and by the Haitian data discussed by Lefebvre (1998a). Second, PCs formed on the basis of substrata from different language families are expected to reflect the idiosyncrasies of their respective substratum languages. This is also borne out by data drawn from the two studies mentioned above. While Solomons Pidgin reproduces the specific features of the Austronesian languages, Haitian Creole reproduces those of the West African languages (for a preliminary discussion of this issue, see Lefebvre 1996b, 1998a). By the same token, the approach taken here helps clarify another issue raised by Gilbert (2002): "Will the typology (language grouping) of creoles ultimately link them more strongly to their constituent languages, or will it link them more strongly to each other?" While some authors advocate the latter possibility (for example, McWhorter 1998a), the analysis presented in this paper would tend to favor the former. Indeed, PCS are hybrid languages which derive the semantic and syntactic properties of their lexical entries from their substratum languages and the phonological representations of these lexical entries from their superstratum languages. Given that they may be formed from different substratum and superstratum languages, what unites them is not their actual features, but rather the processes by which they are formed and the fact that they all emerge in language contact situations. (For further discussion of this issue, see also chapter 8 of this book.)

36

ISSUES IN THE STUDY OF PIDGIN AND CREOLE LANGUAGES

will be adopted, researchers must bear in mind the fact that pidgins and creoles constitute a problem for the genealogical reconstruction of language families. Furthermore, if relexification is the main process at work in the creation of these languages, and since relexification is a cognitive process, it cannot be excluded that in the course of history, several PCs were created without our knowing it!

CHAPTER 3

THE RELEXIFICATION ACCOUNT OF CREOLE GENESIS THE CASE OF HAITIAN CREOLE*

This chapter consists of a summary of my book entitled Creole Genesis and the Acquisition of Grammar: the Case of Haitian Creole published by Cambridge University Press in 1998. This book summarises 25 years of funded research at UQAM1 on the problem of creole genesis. It provides an account of the genesis of creole languages cast within the framework of the processes otherwise known to play a role in the formation of new languages and in language change in general. Three major processes are considered: relexification, reanalysis and dialect levelling. The idea that these processes play a role in creole genesis is not new. The contribution of our research has been to provide a clear statement of how the superstratum data are processed in relexification, of how relexification applies in the case of functional category lexical entries and derivational affixes, and of how word order is established in creole genesis. Furthermore, the contribution of our research has been to develop a theory of how these three processes interact in the formation and development of creole languages, and to document in detail their relative importance. Of these processes, relexification is basic as the two others are hypothesised to apply to the output of relexification. Relexification is also the central process in creole genesis as it accounts for the bulk of the properties of a radical Creole's lexicon: creole lexical entries have phonological representations that are derived from their superstratum languages, they have semantic and syntactic properties that are derived from their substratum languages (see e.g. Adam 1883: 47; Alleyne 1966, 1980; Goodman 1964;Huttar 1975: 684; Sylvain 1936: 178; Voorhoeve 1973; etc.). Reanalysis is much less important than has generally been assumed when creoles' substratum languages are taken into account. Dialect levelling is an important process, but more research needs to be done before a precise characterisation of its weight can be arrived at. The test of our account of creole genesis was based on an extensive and detailed study of Haitian Creole and of its contributing languages: French, its superstratum language, and Fongbe, one of its substratum languages. We were able to gather the resources needed to test this hypothesis * I would like to thank Bernard Comrie for his comments on an earlier draft of this chapter, Edward Raasch for his help in editing it and Andrée Bélanger for formatting it. 1 The history of this research is summarised in the Preface to Lefebvre (1998a).

38

ISSUES IN THE STUDY OF PIDGIN AND CREOLE LANGUAGES

from a global perspective. To my knowledge, this was the first time in the history of the field that such a large enterprise has been undertaken. Section 3.1 presents the relexification hypothesis of creole genesis and section 3.2 the methodology developed for testing it. Section 3.3 summarises the data and conclusions of the extensive comparison of the lexicons of Haitian Creole and of its contributing languages. The conclusion to section 3.3 is that the bulk of the Haitian Creole lexicon has been created through relexification. Section 3.4 summarises the findings pertaining to parameter settings in the formation of Haitian Creole. It is shown that with one exception, Hatian Creole has the parametric values of its substratum languages. To a large extent, this is due to the fact that the functional category lexical entries of the substratum languages have been reproduced in the creole through relexification. Furthermore, in the one case where Haitian Creole does not have the parametric value of its substratum languages, it does not have that of its superstratum language either. Section 3.4 concludes the chapter with a summary of the consequences of the findings. Appendix 2 provides an overview of the Haitian lexicon by class of lexemes with respect to origin. The phonemic inventories of Haitian, French and Fongbe, the correspondences between French phonetic matrices and Haitian phonological forms, as well as orthographic conventions can be found in Lefebvre (1998a: 398-403). Tones on Fongbe words are phonemic. 3 . 1 . The relexification hypothesis of creole genesis The basic hypothesis set forth by this research is that the process of relexification plays a central role in the formation of pidgin and creole languages. Two other processes, reanalysis and dialect levelling, also play a role in the further development of the creole. This section begins with the definition of the cognitive process of relexification followed by the presentation of the relexification hypothesis of creole genesis. The interaction between relexification and the other two processes that play a role in the development of creoles is then presented. Our hypotheses on how functional categories get relexified and on how word order gets established in creolisation are then presented. The section ends with an evaluation of this account of creole genesis. 3.1.1 The process of relexification The first formal definition of relexification was provided by Muysken (1981a: 61): "Given the concept of lexical entry, relexification can be defined as the process of vocabulary substitution in which the only information adopted from the target language in the lexical entry is the phonological representation." Muysken's representation of the process is reproduced in (1).

THE RELEXIFICATION ACCOUNT OF CREOLE GENESIS

39

As per the representation in (1), relexification is a mental process that builds new lexical entries by copying the lexical entries of an already established lexicon and replacing their phonological representations with representations derived from another language. Lefebvre and Lumsden (1994a, 1994b) refer to this second phase as relabelling. According to Muysken's (1981a: 62) proposal, relexification is semantically driven. "For relexification to occur, the semantic representations of source and target language entries must partially overlap; otherwise, the two entries would never be associated with each other. Other features of the two entries may, but need not, be associated with each other." Muysken's representation of relexification was formulated on the basis of data drawn from mixed languages and more particularly on the basis of Media Lengua, a mixed language spoken in Ecuador. In contexts where mixed languages emerge, the speakers who relexify their lexicons are claimed to be bilingual, that is, to master both the source and the target language (see e.g. the papers in Bakker and Mous (eds) 1994). In contrast, in situations where creole languages emerge, speakers of the source languages do not have adequate access to the phonological representation nor to the other properties of the lexical entries of the target language, in this case, the superstratum language. In order to accommodate the representation of the process to these situations, Lefebvre and Lumsden (1994a, 1994b) propose a slightly different representation of relexification reproduced in (2).

40

ISSUES IN THE STUDY OF PIDGIN AND CREOLE LANGUAGES

In the above representation, relabelling proceeds on the basis of phonetic strings found in the superstratum language rather than on the basis of the phonological representations of the lexifier lexical entries. This representation accommodates the numerous cases like that of Tok Pisin baimbai 'later' derived from the English expression by and by (Sankoff and Laberge 1973). Furthermore, the phonetic strings of the lexifier language are interpreted by the relexifiers on the basis of their own phonological system such that the phonological form of the new lexical entry is often quite different from the lexifier language form (Brousseau in preparation). Although the phonological system of a creole appears to be historically derivable from that of its substratum languages, the resulting system is still distinct from the substratum systems. The lexical entry created by relexification in (2) thus has a phonological representation which differs from those of both of its source languages, a fact that is represented by j ' identifying the phonological representation of the new lexical entry in (2). A second point of difference from Muysken's representation is that the lexifier language lexical entry in (2) is deprived of features (compare (1) and (2)). This is due to the fact that, in creole genesis, relexifiers do not have access to the feature of the target language lexical entry. For example, the Haitian verb bezwen 'to need' takes its phonological representation from the French noun besoin 'need'. However, Muysken's insistence on partial semantic overlap between the source and target lexical entries is preserved in the representation in (2) by specifying that the meaning of the phonetic string selected to relabel a copied lexical entry is deduced from its use in specific semantic and pragmatic contexts. Relexification, as represented in (1) and (2), thus consists in copying the properties of a lexical entry and relabelling it. In Lefebvre (1998a), it is assumed that copying applies to all lexical entries and that it is relabelling which is semantically driven. Thus, only those functional categories which have some

THE RELEXIFICATION ACCOUNT OF CREOLE GENESIS

41

semantic content (e.g. determiners, demonstrative terms, etc.) may be assigned a new label during relexification. Those functional categories which have no semantic content (e.g. case markers, operators, etc.) are copied but not relabelled. They are assigned a phonologically null form at relabelling. This null form is represented as [0] in (2). Lexical entries that have a phonologically null representation are not pronounced, nonetheless, they can be argued to be syntactically pertinent. Mous (1995: 1) has yet another representation of the process that he refers to as paralexification. As Mous puts it: "Paralexification is the addition of a word form to a lexical entry. This added form is on par with the existing word form of the lexical entry in question. That is: two word forms share meaning, metaphorical extensions, and morphological properties such as noun class membership for nouns and predicate frame for verbs." Mous's definition of paralexification can be schematised as in (3), where a given lexical entry has two phonological representations and only one set of semantic and syntactic features.

Paralexification and relexification may be viewed as two slightly different ways of representing the same cognitive process. Both representations describe a process which consists in creating a new phonological representation for an already established lexical entry. Both representations allow for the availability of both forms in the competence of speakers over (a certain period of) time. Indeed, based on the representation in (1) and (2), speakers have two parallel lexicons, the original one, and the one created by relexification, which they can use alternately. In the representation in (3), however, speakers have a single lexicon wherein each lexical entry has two phonological representations which can be used alternately. The representations in (1) and (2) can easily be recast within Mous's framework provided that paralexification is considered to be semantically driven, and that it allows for phonologically null forms. Under the representation in (3), relexification can be seen as the addition of a phonological representation to a given lexical entry and of the subsequent loss of the original phonological representation yielding the representation in (4).

The representations in (3) and (4) have the advantage of abstracting the process of relexification away from the social context in which it occurs.

42

ISSUES IN THE STUDY OF PIDGIN AND CREOLE LANGUAGES

In spite of the differences between them, under all three representations, the lexical entries produced by relexification have the semantic and syntactic properties of those in the original lexicons; they differ from the original entries only in their phonological representations. (For further discussion on the representation of relexification, see Lefebvre 1998a: 15-19, 22, 27, 384-6). 2 3.1.2 The hypothesis The hypothesis of our research3 is that the creators of a creole language, adult native speakers of various languages, use the properties of their native lexicons, the parametric values and the semantic interpretation rules of their native grammars in creating a creole. The bulk of a Creole's lexical entries is created by the process of relexification. Two other processes, fed by the output of relexification, dialect levelling and reanalysis, also play a role in the development of a creole (see below). It is claimed that, in creole genesis, the process of relexification is used by speakers of the substratum languages as the main tool for acquiring a second language, the superstratum language. As is pointed out in Lefebvre and Lumsden (1994a), the relexification hypothesis of creole genesis is a further development of the second language acquisition theory of creole genesis. For example, Alleyne (1971, 1980), Andersen (1980), Mufwene (1990), Schumann (1978), Thomason and Kaufman (1991), Valdman (1980), etc. have proposed that pidgin/creole languages constitute a crystallised incomplete stage of second language acquisition. Indeed, in our approach, it is claimed that, in creole genesis involving situations where there is limited access to the superstratum language, the process of relexification is used by speakers of the substratum languages as the main tool for acquiring the superstratum language (for further discussion, see Lefebvre 1998a: 9-12). As is pointed out in Lefebvre (1998a: 10), however, without relexification, the second language acquisition approach to creole genesis does not explain why creole languages have crystallised in the way they have. The relexification hypothesis does explain why creole lexicons reflect the properties of both their superstratum and substratum source languages in the way they do.

2

The relationship between transfer, calquing and relexification is discussed in Lefebvre (1998a: 33-35). The literature bearing on the contexts where relexification may apply (e.g. in the formation of mixed languages, of pidgin languages and in second language acquisition) is reviewed in Lefebvre (1998a: 19-41). The differences between mixed and pidgin/creole languages are discussed in Lefebvre (1998a: 29-30). 3 See Lefebvre 1986, 1993a; Lefebvre and Kaye (eds) 1986; Lefebvre and Lumsden 1989, 1994a, 1994b.

THE RELEXIFICATION ACCOUNT OF CREOLE GENESIS

43

3.1.3 The interplay of relexification, reanalysis and dialect levelling in creole genesis and development As was mentioned above, in addition to relexification, two other processes were hypothesised to be involved in the development of a creole: dialect levelling and reanalysis. Dialect levelling, as discussed in the literature on dialects in contact (e.g. Domingue 1980; Trudgill 1986; Siegel 1995), refers to the reduction of variation between dialects of the same language in situations where these dialects are brought together. Reanalysis, a major process in language change, is a mental process whereby a particular form which signals one lexical entry becomes the signal of another lexical entry (e.g. Lightfoot 1979). The content of this section summarises the interplay of these three processes as they are embedded within the scenario of creole genesis developed in Lefebvre and Lumsden (1989, 1992, 1994a, 1994b) and in Lumsden and Lefebvre (1994). In this scenario, relexification applies in creole genesis in the following way. Native speakers of various substratum languages are brought together. Crucially, the speakers of the substratum community do not have a common language, a situation which creates the need for a lingua franca not only to communicate with the colonisers but also to communicate among themselves. The substratum speakers are exposed to a superstratum language, the language of the colonists. However, they do not have enough exposure to this language to learn the details of its lexical entries. Due to this situation, speakers of the substratum languages relexify the lexical entries of their respective lexicons on the basis of phonetic strings found in the superstratum language (see (2)). The relexification of various lexicons on the basis of a single superstratum language provides the speakers of the substratum languages with a common vocabulary. As is pointed out in Lefebvre and Lumsden (1994a, 1994b), it is the limited direct access to the superstratum language that makes relexification so important in the formation of radical4 creoles. It is a well documented fact that in creoles, both functional category items as well as major category lexical entries have phonological representations that are similar to some phonetic strings of the superstratum language. These lexical entries, however, do not have the same properties as the corresponding superstratum forms from which they are phonologically derived (see e.g. Carden and Stewart 1988; Lefebvre 1984; Lefebvre and Lumsden 1989, 1992; Mufwene 1991). In the scenario of creole genesis reported on here, it is hypothesised that because speakers of the substratum languages have very limited access to the superstratum data, they typically fail to identify the functional categories of the superstratum language. These speakers thus try to 4 Creoles which present little similarity with their superstratum languages are referred to as radical creoles.

44

ISSUES IN THE STUDY OF PIDGIN AND CREOLE LANGUAGES

relexify the functional items of their native languages on the basis of forms found in the superstratum language. It is proposed that the functional category lexical entries of the substratum languages are relexified on the basis of major category lexical items (e.g. nouns, adjectives, verbs, adverbs and prepositions) of the superstratum language with which they share some semantics and distributional properties (for further discussion, see Lefebvre 1998a: 35-41). For example, the postposed definite determiner of the substratum languages of Haitian Creole is argued to have been relexified on the basis of the postposed French adverb là, yielding Haitian la (see Lefebvre 1998a: 79-84, and section 3.3.4.1). The relexification of functional as well as major category lexical entries provides the speakers of the various substratum languages with a common vocabulary in all areas of the lexicon. As we saw in section 3.1.1, relabelling is semantically driven, in the sense that there must be partial semantic overlap between the source and target lexical entries for it to take place. Consequently, relabelling is constrained by what the superstratum language has to offer in terms of appropriate strings to relexify original lexical entries. This is particularly crucial in the case of functional category items. It is thus possible that some lexical entries cannot be assigned a new phonological form because there is no form available in the superstratum language to provide a new phonological form for particular lexical entries. In this case, the new lexical entry is assigned a phonologically null form. As we saw above, phonologically null lexical entries in the creole may also arise from the fact that the original lexical entry has no semantics (e.g. operators, case markers, etc.). The lexicons created by relexification become the basis of a lingua franca within the creole community. When the relexified lexicons become the target of the creole community, a new language is born. At this point, the speakers are no longer targeting the superstratum language. They are now targeting the common language that they have developed through relexification: the incipient creole. At this stage, two other processes come into play: dialect levelling and reanalysis. Relexification is a cognitive, hence an individual process. Situations where creoles emerge involve several substratum languages. Each individual relexifies his or her own lexicon. Speakers of various substratum languages reproduce the idiosyncratic semantic and syntactic properties of their own lexicons in relexification and thus, the product of relexification is not necessarily uniform across the creole community. For example, relexification of the lexicons of languages X, Y and Z on the basis of a single superstratum language will yield three slightly different lexicons in an incipient creole. This is schematically represented in (5).

THE RELEXIFICATION ACCOUNT OF CREOLE GENESIS

45

The relexification of several lexicons thus creates variation within a creole. In our scenario of creole genesis, dialect levelling is hypothesised to apply to the output of relexification in order to reduce the variation produced by the relexification of the various substratum lexicons. This scenario allows for a sound explanation of the facts referred to in the literature on creole studies as the 'cafeteria principle'—a term used first by Dillard (1970) and later by Bickerton. As Bickerton (1981: 49) puts it: "As things stand, we are asked to believe that different African languages contributed different rules and features to particular creoles (...) it is (...) absurd to suppose that a creole could mix fragments of Yoruba, Akan, Igbo, Mandinka, and Wolof (...)." The proposal that dialect levelling operates on the output of the various relexified lexicons involved in creole formation provides a principled explanation of the observation that several different substratum languages may contribute features to a given creole. Plural forms, reflexive forms, demonstrative terms and the imperfective constructions, to name but a few, are cases in point. (See the data sections of this chapter and chapter 9; see also Lefebvre (1998a) for the discussion of more cases.) In the competition among different creole dialects (created by the relexification of different substratum lexicons), there are winners and losers. As is discussed in Lefebvre (1998a: 390-391), the competition is not always won by speakers of the same relexified lexicon (see also Siegel 1997). An original lexical entry that was not assigned a label during relexification, either because it had no semantic content and thus could not be relabelled, or because there was no appropriate form in the superstratum language to relabel the copied lexical entry, may be signalled by a periphrastic expression. For example, a given tense or aspect may be signalled by an adverb with a similar meaning. The periphrastic expression may later become the phonological representation of the functional category in question through the process of reanalysis. A case in point in Haitian is the reanalysis of the sentence initial adverb of posteriority apre as the marker of definite future ap (see section 3.3.5.1). This case is similar to the more renowned Tok Pisin case: the sentence initial adverb of posteriority baimbai reanalysed as the perverbal marker bai (see Sankoff 1991). In both cases, the lexical entry that has become overt through reanalysis has the properties of the corresponding lexical entries in the substratum languages since the creole lexical entry has been created by relexification, though without having been relabelled. The postulated link between relexification and reanalysis accounts in a straightforward way for the

46

ISSUES IN THE STUDY OF PIDGIN AND CREOLE LANGUAGES

paradoxical situation noted in the literature according to which, in the course of their further development, creoles develop lexical entries that manifest the properties of their substratum languages even in situations where the substratum languages have ceased to be spoken (see e.g. Chaudenson 1994; Mufwene 1990; Mühlhäusler 1986a, 1986b; Sankoff 1991: 73). (For an extensive discussion of this point, see Lefebvre 1998a: 108-110, 375-386, and the references therein.) (Further discussions of dialect levelling and reanalysis and of their interaction with relexification in creole genesis may be found in Lefebvre 1998a: 41-47. See also chapter 9 of this book.) 3.1.4 Word order The problem of how word order is established in creole genesis has been raised several times in the literature (e.g. Mühlhäusler 1986b: 47; Mufwene 1990: 5). Lefebvre and Lumsden (1992) make a twofold proposal. First, because the relexifiers intend to reproduce the phonetic strings of the superstratum language—an assumption that follows logically from the claim that creole genesis is a function of second language acquisition—and because they identify major category lexical items of the superstratum language, they acquire the directionality properties of the superstratum major category lexical entries. Thus, the word order of major category lexical entries in the creole is predicted to follow the word order of lexical categories in the superstratum language. Consequently, if the superstratum language has prenominal adjectives, the creole will have prenominal adjectives. Likewise, creoles whose lexifier languages have prepositions but no postpositions are predicted to have only prepositions. Second, because they do not have enough access to the superstratum language, the creators of a radical creole do not identify the functional categories of that language. In Lefebvre and Lumsden (1992), it is hypothesised that the creators of the creole retain the directionality properties of the functional category lexical entries of their own lexicon in relexification. Consequently, the creole functional categories will have the same word order as the substratum entries that they were copied from. To a large extent, this proposal is borne out, as we will see in the data sections of this chapter (for further discussion, see Lefebvre 1998a: 38-40, 89, 180, 388-390). 3.1.5 An optimal account of creole genesis The theory of creole genesis outlined in this section provides a straightforward and optimal account of the properties of creole languages. The following discussion builds on a preliminary one in Lefebvre and Lumsden (1989, 1994a). First, by virtue of the definition of the process, creole lexical entries are predicted to have the same semantic and syntactic properties as the corresponding lexical entries in the substratum languages, but phonological

THE RELEXIFICATION ACCOUNT OF CREOLE GENESIS

47

representations derived from the phonetic strings of the superstratum language. The hypothesis thus explains why creoles reflect the properties of both their superstratum and their substratum source languages in the way they do (e.g. Sylvain 1936; Goodman 1964; Hurtar 1975; Keesing 1988; etc.). As was pointed out in Lefebvre and Lumsden (1989), the hypothesis that relexification plays a central role in creole genesis is falsifiable. If a comparison of the lexical properties of a radical creole with the lexical properties of its source languages were to show that the syntax and semantics of the creole are not systematically parallel to the syntax and semantics of the substratum languages, then the hypothesis would be falsified. Likewise, the hypothesis that the creators of the creole use the parametric values, semantic interpretation rules and principles of concatenation of their own grammars in creating the creole explains why creoles coincide with their substratum languages in these areas of the grammar as well. Again, this hypothesis is falsifiable. If a comparison of the grammatical and semantic properties of a radical creole with those of its source languages were to show that the properties of the creole are not systematically parallel to those of the substratum languages, then the hypothesis would be falsified. The second main point is that, given the multilingual situation prevailing in contexts where creoles emerge (e.g. Whinnom 1971), and given the urgent need for a lingua franca in such a situation (e.g. Hymes 1971; Foley 1988; Thomason and Kaufman 1991; etc.) and, furthermore, given the limited access to the superstratum language (e.g. Thomason and Kaufman 1991; Foley 1988; etc.), only one generation of speakers is required to form a new language by means of relexification and the use of the parametric values and other principles of the native grammars. The claim that these processes are at work in creole genesis accounts for the fact that creole languages can be created relatively quickly as compared with regular cases of linguistic change (e.g. Voorhoeve 1973; Hancock 1987; etc.). Finally, as is observed in Lefebvre and Lumsden (1994a), the fact that creoles are generally isolating languages can also be deduced from the above proposal. Since the minor category lexical entries of creole languages derive their phonological forms from major category lexical items in the superstratum language, or from reanalysis, and since these categories are typically free morphemes, it follows that creoles will tend to be isolating languages. 3.1.6 Summary The account of creole genesis presented in this section can be reduced to three major processes that interact in a specific way: relexification, a central process in language genesis, and reanalysis and dialect levelling, which apply to the output of relexification. The following section discusses the methodology that was

48

ISSUES IN THE STUDY OF PIDGIN AND CREOLE LANGUAGES

developed in order to test this hypothesis. The evaluation of the hypothesis against the data is presented in the remaining sections. 3 . 2 . The test of the hypothesis The hypothesis was tested using Haitian Creole. The research programme invol ved two dimensions, historical and linguistic. These will be discussed in turn. 3.2.1 The historical research The historical research was designed to answer the following questions: When was Haitian Creole formed? What were the salient demographic characteristics of the Haitian population during that period? Who were the people present at the relevant time? What was their linguistic background? (see Lefebvre 1993a). The historical research in the colonial archives of France was carried out by John Singler (see Singler 1993a, 1993b, 1996). In short, Singler establishes the following points. Haitian Creole was formed between 1680 and 1740. This period is characterised by the following features. As a consequence of a shift from a tobacco and cotton economy to a sugar economy, the number of colonists decreased and the number of slaves exploded; this had the effect of modifying the slave population's exposure to French (Singler 1996). The bulk of the Caribbean population at the time Haitian Creole was formed was adult. As for the languages that these adults were speaking, Singler (1993b) shows that they were all Niger-Congo languages, more particularly Kwa (Gbe and Akan) and Bantu. During the formative period of Haitian Creole, Gbe speakers made up more than 50% of the French Caribbean slave-export population. As is pointed out in Lefebvre and Lumsden (1994b), the overall situation found in Haiti between 1680 and 1740 presented all the prerequisites for the emergence of a creole language: there was a multilingual community, in need of a lingua franca, and the bulk of the population, the speakers of the substratum languages, had only reduced access to the superstratum language (for further discussion, see Lefebvre 1998a: 52-58). 3.2.2 The linguistic test The linguistic test consists in a detailed comparison of the lexicon and grammar of Haitian Creole with those of its contributing languages: French, its superstratum language, and West African languages, its substratum languages. Due to time and resource constraints, the detailed study of the substratum languages of Haitian was limited to one language. Because of the importance of the influence of the Fon culture on that of Haiti (with respect to religion and art, see e.g. Bastide 1967; Herskovits 1975), Fongbe, a language of the Gbe cluster, was chosen as the substratum language to be studied in detail (see Lefebvre 1986, 1993a; Lefebvre and Kaye (eds) 1986). Note that in no way does this

THE RELEXIFICATION ACCOUNT OF CREOLE GENESIS

49

methodological choice entail that the formation of Haitian can be reduced to the relexification of Fongbe alone. Our choice of Fongbe turned out to be a good one, in view of Singler's finding that the Gbe speakers outnumbered speakers of the other West African languages at the time Haitian Creole was formed. Furthermore, as is discussed at length in Lefebvre (1998a: 58-62), although the African languages spoken in Haiti at the time Haitian Creole was formed were numerous, they share a significant number of typological properties such that they constitute a relatively homogeneous group. As has been pointed out in Lefebvre and Lumsden (1994a), the methodological choices that we had to make had the effect of making the relexification hypothesis easier to falsify. (For a thorough discussion of the methodology of the research and the validity of the linguistic test, see Lefebvre 1998a: 52-77, and the references cited therein.) The linguistic test involves a global comparison of the lexicons, parametric values, semantic interpretation rules and concatenation principles of the languages involved. As has been emphasised in Lefebvre and Lumsden (1994a, 1994b), one or two examples either way are not enough to support or falsify the hypothesis.5 The test must involve quantity as well as quality. The comparison of the three lexicons is summarised in section 3.3 and that of the parametric options in section 3.4. The conclusions to the detailed analyses layed out in the book are simply stated here, and the reader is referred to the various sections of the book for the analyses themselves. 3 . 3 . The lexicon This section summarises the findings pertaining to the role of relexification, reanalysis, and levelling in the formation of the Haitian Creole lexicon. Section 3.3.1 is dedicated to lexical semantics, section 3.3.2 to the syntactic properties of verbs, section 3.3.3. to derivational affixes. Sections 3.3.4 and 3.3.5 discuss functional category lexical entries involved in nominal structure and in clause structure, respectively. Rules and principles of concatenation of morphemes and lexemes will be mentioned throughout this section whenever pertinent. 3.3.1 Lexical semantics This section brings together data illustrating the fact that, while the forms of the Haitian major category lexical items are derived from French, the semantic properties of these Haitian words are derived from the substratum languages. Nouns, pronouns, reflexives, wh-words and verbs will be discussed in turn. 5 A thorough discussion of the research methodology can be found in Lefebvre (1998a: 52-78) where the following methodological points are raised: the typological features of the source languages of Haitian, the superstratum data the creators of Haitian were exposed to, the linguis tic test designed to test the relexification hypothesis of creole genesis, what counts as evidence for the hypothesis and how it can be falsified, the source of data and the mode of data analysis.

50

ISSUES IN THE STUDY OF PIDGIN AND CREOLE LANGUAGES

3.3.1.1 Simplex and compound nouns. Consider the nouns in (6). The Haitian lexical entries all have two meanings. For example, the noun plim means both 'feather' and 'hair'. Its form is derived from that of the corresponding French lexical entry plume. However, the French lexical entry has only one meaning, and therefore it cannot be the source of the extra meaning associated with the Haitian lexical entry. The corresponding Fongbe lexical entry, however, has the same two meanings as the Haitian one. This shows that the substratum lexical entry is the source of the semantic properties of the Haitian entry. The nature of the process of relexification predicts the properties of the Haitian lexical entries in (6): these lexical entries have a phonological representation derived from French but semantic properties derived from the substratum language. (6)

HAITIAN

FRENCH

FONGBE

plim 'feather' 'hair' vyann 'meat' 'edible animals' (complement of the verb 'to kill') dife 'fire' 'brand' tèt 'head' 'roof' van 'wind' 'air'

plume 'feather'

fún

viande 'meat'

làn

{du) feu 'fire' tête 'head' vent 'wind'

'feather' 'hair' 'meat' 'edible animals' (complement of the verb 'to kill') myòn 'fire' 'brand' ta 'head' 'roof' jòhòn 'wind' 'air' (from Lefebvre 1998a: 71)

The process of relexification also explains why some French lexical entries have not made their way into Haitian Creole. For example, as is shown in (7), while French BODY-parts are referred to by means of simplexes, Haitian BODY-parts are referred to by means of compounds. The words that are compounded are all phonologically derived from French but the French Simplexes refering to BODY-parts did not make their way into Haitian. As is shown in (7), BODY-parts in Fongbe are referred by means of compounds. On the one hand, the data show that the Haitian compounds are formed on the model of the Fongbe ones. (For a discussion on the ordering of words in Haitian and Fongbe compounds, see Lefebvre 1998a: 339-342, and the references therein.) On the other hand, the distribution in (7) suggests that the French simplexes

THE RELEXIFICATION ACCOUNT OF CREOLE GENESIS

51

referring to BODY-parts did not make their way into Haitian because the creators of Haitian did not have simplexes to relexify in these cases. (7)

FRENCH

HAITIAN

FONGBE

lèvre 'lip' narine 'nostril' cil 'eyelash' nuque 'nape' crâne 'skull'

po-bouch nù-fló 'lip' 'skin-mouth' 'mouth-skin' twou-ne àcntín-dó 'nostril' 'hole-nose' 'nose-hole' plim-je wùn-dà 'eyelash' 'hair-eye' 'eye-hair' dèyè-kou kc-gùdó 'nape' 'back-neck' 'neck-back' kalbas-tèt tà-kâ 'skull' 'calabash-head' 'head-calabash' or tèt-kalbas (from Brousseau 1989) In a similar fashion, in (8), where French has simplexes referring to people having certain characteristics, Haitian has compounds referring to people having the same characteristics. Again, the Haitian compounds are built on the model of the substratum language. (8)

FRENCH

HAITIAN

FONGBE

a. chauve b. c. d. e. f.

tèt-chöv tà-súnsún 'bald (person)' {tête-chauve) 'head-bald' 'head-bald' personne- tèt-chaje tà-gbà 'problematic problème (tête-trouble) (person)' 'head-troubled' 'head-troubled' aveugle je-pete nukún-tón-no 'blind (person)' (oeil-pété) 'eye-burst' 'eye-burst-ATT' audacieux je-chèch hcn-wùn 'audacious (person)' (oeil-sec) 'eye-dry' 'clear-eye' prétentieux je-fô nùkun-kέn 'pretentious (person)' (oeil-fort) 'eye-strong' 'eye-strong' entêté tèt-di tà-mέ-siέn-tc 'stubborn (person)' or tête-dure (tête-dure) 'head-hard' 'head-in-hard-AG' (from Lefebvre 1998a: 336)

The fact that the French simplexes in (8) did not make their way into Haitian is accounted for by the relexification hypothesis.

52

ISSUES IN THE STUDY OF PIDGIN AND CREOLE LANGUAGES

The data in (7) and (8) further show that concepts that are rendered as compounds in Fongbe are often also rendered as compounds in Haitian where French has Simplexes. Furthermore, in Haitian, simplexes are compounded following the semantics of the substratum language rather than the superstratum language. These facts argue for the claim that the creators of the creole used the principles of their own grammar in concatenating simplexes. Finally, as is shown in Brousseau (1988, 1989), with the exception of synthetic compounds, the types of compounds found in Haitian parallel the types in Fongbe rather than in French. (For further discussion, see Lefebvre 1998a: 334-348.) 3.3.1.2 Pronouns. The paradigms of tonic personal pronouns, possessive adjectives and pronouns, logophoric pronouns and expletives will be discussed in turn. 3.3.1.2.1 Personal pronouns. The paradigm of French tonic pronouns is as in (9). Gender (masculine and feminine) is distinguised in the third person singular and plural. (9)

moi toi lui/elle nous vous eux/elles

1st person singular 2nd person singular 3rd person singular 1st person plural 2nd person plural 3rd person plural

T, me' FRENCH 'you (sg)' 'he (m)/she (f)' 'we/us' 'you (pl)' 'they, them (m)/they, them (f)'

The paradigm of Haitian tonic pronouns in (10) has forms that are all derived from the French ones in (9). Gender is not encoded in this paradigm. The most striking fact about this paradigm is that the same form is used for both first and second person plural. This contrast with French, which has two distinct forms to encode first and second person plural. (10)

mwen ou/[wu] li nou yo

1st person singular 2nd person singular 3rd person singular 1st and 2nd person plural 3rd person plural

T, me' HAITIAN 'you (sg)' 'he/she/it/him/her' 'we/us/you (pl)' 'they/them' (from Valdman et al. 1981)

In the Fongbe paradigm of personal pronouns in (11), there is no gender distinction. Interestingly enough, the same form is used to encode both first and second person plural (see Lefebvre and Brousseau 2002: 61).

THE RELEXIFICATION ACCOUNT OF CREOLE GENESIS

(11)

nyέ h wέ é(yέ) mí yé

53

1st person singular 'I, me' FONGBE 2nd person singular 'you (sg)' 3rd person singular 'he/she/it/him/her' 1st and 2nd person plural 'we/us/you (pl)' 3rd person plural 'they/them' (=(18) in Brousseau 1995a)

The discrepancies between the Haitian and the French paradigms of personal pronouns appear to come from the substratum language and thus to follow in a straightforward way from the relexification hypothesis. The morpheme yo is further discussed in section 3.3.4.2. Furthermore, the distribution of the tonic pronouns is not the same in all three languages. In Fongbe, tonic pronouns participate in the possessive cons truction. In this construction, the possessed noun is followed by the possessor; the latter can surface either as a noun or as a tonic pronoun followed by the genitive case marker, as is shown in (12). (The analysis according to which ton is the genitive case marker in (12) is due to Brousseau and Lumsden 1992). (12)

a. b.

[xwé [Bàyí tòn]] FONGBE house Bayi GEN 'Bayi's house' [xwé [nyè/ hwe/ é / mí I yé ton]] FONGBE house me / you / (s)he / we, you / they GEN 'my/your/his, her/our, your/their house' (=(30) in Brousseau 1995a)

In contrast, French pronouns do not occur in this context: *maison à/de moi (lit.: 'house of me') is impossible in all French dialects (see Lefebvre 1998a: 143-147 for further discussion). Haitian follows the Fongbe pattern in using strong pronouns in possessive constructions, as is exemplified in (13). Note, however, that the case marker following the possessor in (13) is phonologically null. As per the analysis in Lumsden (1991), this null form has the properties of genitive case (see also Lefebvre 1998a: 101-110). (13)

a. b.

[kay [Jan 0]] HAITIAN house John GEN 'John's house' (from (8a) in Lefebvre 1998a: 145) [kay [mwen / u / ...ø]] HAITIAN house me / you / ... GEN 'my/your house' (from (8b) in Lefebvre 1998a: 145)

Fongbe also makes use of tonic pronouns in possessive constructions where the possessed is not specified. This is illustrated in (14) where the non-

54

ISSUES IN THE STUDY OF PIDGIN AND CREOLE LANGUAGES

specified possessed NP is followed by the possessor which can be expressed either as a noun or as a personal pronoun marked for genitive case. (14) Kdkú sc [— [Àsíbá I nyè tcn]] Koku take Asiba me GEN 'Koku took Asiba's/mine.'

FONGBE (=(11) in Lefebvre 1998a: 146)

French tonic pronouns cannot appear in this type of construction, as the sentence *Koku a pris de moi (lit.: 'Koku took of me') is not licit. Again, Haitian follows the Fongbe pattern as it encodes reference to a whole possessive NP by means of the genitive construction involving a noun or a personal pronoun followed by the phonologically null genitive case marker, as is illustrated in (15). (15)

Jan pran ¡pa [Mari Imwen John take thing Mary me GEN 'John took Mary's/mine.'

ø]]

HAITIAN

(=(12) in Lefebvre 1998a: 146)

The Haitian structure in (15) is of the same type as the Fongbe one in (14), except for the fact that the possessed NP in Haitian is realised as pa, a head filler. Haitian pa is phonologically derived from the French form part 'share' (Goodman 1964). However, it is most appropriately glossed as 'thing' since it may refer to any possible antecedent that may be possessed. This contrasts with Fongbe, where the possessed NP is phonologically null. The difference between the two languages thus appears to be that, while Fongbe allows the possessed to be null, Haitian does not (Brousseau 1995a; Kinyalolo 1994). This discrepancy may be attributable to the fact that, while in Fongbe, genitive case is phonologically overt, in Haitian it is not. (For further discussion see Lefebvre 1998a: 143-147). So far, we see that the Haitian paradigm of personal pronouns follows the semantic division of the substratum language in having five pronominal forms instead of six. We also see that the distribution of these Haitian pronominal forms follows that of the Fongbe tonic pronouns and departs in a significant way from that of the French tonic pronouns. Since French tonic personal pronouns do not occur in possessive constructions, as was shown above, the following questions arise: (i) How does French encode these possessive relationships? (ii) Does Fongbe offer a parallel way of encoding these relationships? And finally, (iii), is the French way of encoding these relationships replicated in Haitian? These questions are addressed in the following section. 3.3.1.2.2 Possessive adjectives and pronouns. As we saw in the preceding section, French tonic personal pronouns do not occur in possessive

THE RELEXIFICATION ACCOUNT OF CREOLE GENESIS

55

constructions. This is because French has paradigms of possessive adjectives and pronouns. These paradigms will be discussed in turn, and compared with Fongbe and Haitian. French has a paradigm of possessive adjectives in which forms are distinguished by gender (masculine and feminine) and by number (singular and plural). This is shown in (16). (16)

Possessed singular masculine feminine 1st sg mon ma 2nd sg ton ta 3rd sg son sa 1st pl notre 2nd pl votre 3rd pl leur

Possessed plural mes tes ses nos vos leurs

FRENCH 'my' 'your' 'his/her' 'our' 'your' 'their'

These possessive adjectives precede the possessed noun: mon livre 'my book', ma table 'my table'. Fongbe has a defective paradigm of possessive adjectives that contains only two terms. As is shown in (17), these possessive adjectives follow the noun.. (17)

xwé cè / tòwè house my / your 'my/your house'

FONGBE (=(31) in Brousseau 1995a)

As we saw in the preceding section (see (12)), the regular way of encoding possession in Fongbe is by means of the genitive construction where the possessor, a noun or a tonic personal pronoun, is followed by the genitive case marker. The Haitian lexicon has no possessive adjectives. As is extensively discussed in Lefebvre (1998a: 143-147), this follows from the relexification hypothesis of creole genesis. The French forms in (16) which have no Fongbe counterparts were not incorporated into Haitian because there were no such entries in the original lexicon to be relexified. Brousseau (1995a) further suggests that the Fongbe forms cè and tôwè in (17) were not relexified for the following reasons: First, the native grammar already offered another regular option for encoding the same relationships, namely personal pronouns used in the genitive construction (see (12)). Second, the fact that the Fongbe paradigm of possessive adjectives was defective probably played a role in leading the creators of Haitian to abandon these two lexical entries in creating the new lexicon. This is a likely explanation in light of the fact that, in some Gbe languages, the paradigm of possessive adjectives is not only defective but

56

ISSUES IN THE STUDY OF PIDGIN AND CREOLE LANGUAGES

lacking all together. Hazoumê's (1990) description of possessive constructions in several Gbe dialects reveals the following similarities and differences between them. First, no Gbe dialect has a complete paradigm of possessive adjectives: Like Fongbe, Gungbe only has two forms corresponding to those in (17). Ajagbe and Gengbe only have a form for first person singular, and Tɔfingbe, for second person singular. Other Gbe dialects such as Xwedagbe have no such forms at all. Second, in all Gbe dialects possession may be expressed by means of a genitive construction of the type illustrated in (12b) for Fongbe. In contexts corresponding to those in (14) and (15), where the possessor is not specified, French has a paradigm of possessive pronouns presented in (18) (from Grevisse 1975; Haase 1975). These pronominal forms encode person, number and gender features. Number and gender features must match those of the antecedent NP. (18)

Possessed singular masculine feminine 1st sg mien mienne 2nd sg tien tienne 3rd sg sien sienne 1st pl nôtre 2nd pl vôtre 3rd pl leur

Possessed plural masculine feminine miens miennes tiens tiennes siens siennes nôtres vôtres leurs

FRENCH 'mine' 'yours' 'his/hers/its' 'ours' 'yours' 'theirs'

Gbe languages do not have possessive pronouns (Hazoumê 1990). As we saw in section 3.3.1.2.1, in Fongbe, the type of possessive construction involving possessive pronouns in French is rendered by means of a tonic personal pronoun followed by the genitive case. The possessed NP is phonologically null. As is shown in (14), Haitian does not have possessive pronouns either. This follows from the relexification account of creole genesis: the creators of Haitian simply had no possessive pronouns to relexify. 3.3.1.2.3 Logophoric pronouns. Some West African languages have a form of personal pronoun which, unlike other pronominal forms, has no independent reference (e.g. Clements 1975; Hagège 1974; Hyman and Comrie 1981; etc.). This kind of pronoun is referred to as logophoric. A logophoric pronoun is morphologically distinct from the other pronouns in a given language and it must have an antecedent in a higher clause. In addition to the personal pronouns in section 3.3.1.2.1, Fongbe has a logophoric pronoun. The form of this pronoun is émî. It is morphologically distinct from the personal pronouns in (11). Furthermore, as is shown by the ungrammaticality of (19), émî has no independent reference.

THE RELEXIFICATION ACCOUNT OF CREOLE GENESIS (19) *Émì hwlá Àsíbá sín gbó LOG hide Asiba OBJ goat '(S)he hid Asiba's goat.'

57 FONGBE

(=(6) in Kinyalolo 1993c)

Émî has to be interpreted based on the context in which it appears. The data in (20) show that, in Fongbe, the antecedent of émî may be second person, as in (20a), or third person, as in (20b) and (20c), but not first person (Kinyalolo 1993b, 1993c). (20) Sìká tùn d yé d à flín d émì hwlá FONGBE Sika know say they say you remember say LOG hide Àsíbá sín gb. Asiba OBJ goat a. 'Sika knows that they said that you¿ remember that you; hid A's goat.' b. 'Sika knows that theyi said that you remember that theyi hid A's goat.' c. 'Sikai knows that they said that you remember that shei hid A's goat.' (=(13) in Kinyalolo 1993c) Haitian does not have a logophoric pronoun. In Lefebvre (1998a: 147-148), it is claimed that this follows from the semantic constraint imposed on the process of relabelling in relexification: since relabelling is semantically driven, and, since logophoric pronouns do not have independent semantic content, it follows that logophoric pronouns will not be relabelled in creolisation. Since there has been no data presented to argue for a phonologically null logophoric pronoun in Haitian, it is logical to assume that the logophoric pronoun of the substratum language has not made its way into the creole. 3.3.1.2.4 Expletives. In all three languages under comparison, the form of the expletive subject is the same as that of the third person singular personal pronoun: li in Haitian, il in French and é in Fongbe, as is shown in (21). (21) a. b. c.

Li bon pou Jan pati. HAITIAN it good COMP John leave 'It is good that John will leave.' Il est bon que Jean parte. FRENCH it AUX good COMP John leave 'It is good that John will leave.' É ny d Kkú ní yì. FONGBE it be.good COMP Koku IRR leave 'It is good that Koku will leave.' (=(36) in Lefebvre 1998a: 157)

58

ISSUES IN THE STUDY OF PIDGIN AND CREOLE LANGUAGES

The Haitian expletive is not always overt (e.g. Koopman 1986; Massam 1989; DeGraff 1992a, 1992d, 1993b, 1994; Déprez 1992a; Vinet 1991; Law 1992; etc.). This is illustrated in (22). (22)

(Li) sanble Jan te malad. it seem John ANT sick 'It seems that John has been sick.'

HAITIAN (=(37) in Lefebvre 1998a: 158)

Koopman (1986), Massam (1989), Vinet (1991) and Déprez (1992a) all proposed that, in addition to an overt expletive subject, Haitian has a phonologically null expletive subject. Not all languages have this option. For example, as is shown by the ungrammaticality of (23), French does not have this option. (For further discussion, see Lefebvre 1998a: 157-159.) (23)

*semble que Jean soit malade seem COMP John be sick

FRENCH (=(38) in Lefebvre 1998a: 158)

Fongbe, however, does have this option, as is illustrated in (24). (24)

(É) dì d Kkú j'àzn. it seem COMP Koku sick 'It seems that Koku is sick.'

FONGBE (=(42) in Lefebvre 1998a: 159)

So, in both Haitian and Fongbe, but not in French, a phonologically null expletive is available. Assuming that a phonologically null expletive constitutes a lexical entry, the creators of Haitian would have reproduced it in the creole lexicon. 3.3.1.3 Reflexives. Haitian lacks an overt morphological reflexive form of the type of self in English. Some nouns and pronouns, however, are involved in the interpretation of reflexivity. In fact, Haitian offers three possibilities for expressing this notion, as is illustrated below. The sentence in (25a) shows that a bare personal pronoun may be assigned a reflexive interpretation. The sentences in (25b) and (25c) show that nouns such as tèt 'head' and kò 'body', followed by a possessor phrase containing a personal pronoun, may also be assigned a reflexive interpretation. (25) a. b.

Mweni wè mweni nan glas la HAITIAN I see me in mirror DEF T saw myself in the mirror.' (=(1a) in Brousseau 1995b) Mi ap touye tèt mweni HAITIAN I DEF.FUT kill head me T will kill myself.' (=(2a) in Brousseau 1995b)

THE RELEXEFICATION ACCOUNT OF CREOLE GENESIS c.

blese ko Zn Lii he hurt body him 'He hurt himself. '

59 HAITIAN

(=(6d) in Brousseau 1995b)

In contrast to Haitian, French does not allow a reflexive interpretation of bare pronominal forms, as is shown by the ungrammaticality of (26a). For a reflexive interpretation to obtain, a syntactic clitic has to be used, for the first and second person, as is shown in (26b); in the third person, the reflexive clitic se has to be used as in (26c). (26)

a. b. c.

*Je vois 1st see

moi me

dans in

le DEF

miroir FRENCH miror (=(48a) in Lefebvre 1998a: 161) Je me vois dans le miroir FRENCH 1st 1st see in DEF miror 'I see myself in the miror.' Ils se voient FRENCH 3rd REF see 'They see themselves.'

In Lefebvre (1998a: 162-166), it is extensively argued that French has played no role in determining the reflexive interpretation of the Haitian personal pronouns as in (25a). This conclusion accords with Muysken and Smith's (1995) observation that the lexifier languages can only play a limited role in the historical derivation of reflexives in creole languages in general. In Lefebvre (1998a: 159-171), it is argued, however, that the patterns in (25) find a straightforward explanation in the substratum languages of Haitian. All Gbe languages encode reflexivisation by means of a personal pronoun + SELF (Hazoumê 1990). In Fongbe, this lexical item is -déè and it has semantic and distributional properties that are similar to those of the English -self (Kinyalolo 1994). For example, in English, pronouns combine with -self (e.g. He washes himself). Likewise, in Fongbe, the strong pronominal forms discussed in section 3.3.1.2.1, but not the syntactic clitics (Kinyalolo 1994; Brousseau 1995a), combine with -déè, as is illustrated in (27). (27)

a.

Ni ná hù nyè--déè i 1sg DEF.FUT kill me-SELF T will kill myself.'

b . Bàyíi mo é-déèi FONGBE Bayi see she-SELF 'Bayi saw herself.' (=(45) in Brousseau 1995a)

In English, the construction pronoun + self is assigned an analysis such as that in (28a), where -self is the head of NP, and the pronoun occurs in the specifier position of the projection (Chomsky 1981). Kinyalolo (1994) proposes to

60

ISSUES IN THE STUDY OF PIDGIN AND CREOLE LANGUAGES

account for the Fongbe construction pronoun + déè in the same way, as is illustrated in (28b).

In Haitian, there is no overt form corresponding to Fongbe déè. Nonetheless, Lefebvre (1998a: 164) proposes that the representation of the Haitian reflexive phrase is as in Fongbe, with the difference that, in Haitian, the head noun of the construction is phonologically null, as is illustrated in (29).

In terms of the relexification hypothesis, this amounts to saying that the lexical entry copied from Fongbe déè was assigned a null form at relabelling.6 In Lefebvre (1998a: 160-167), it is hypothesised that this is because the superstratum language did not offer a form that could provide a new label for the original lexical entry. As we saw above, French does not have a lexical anaphor. In this view, then, the Haitian lexicon would have a lexical entry corresponding to Fongbe déè with a phonologically null representation. In this analysis, a personal pronoun is assigned a reflexive interpretation when it is interpreted from the specifier position of an NP headed by a phonologically null reflexive anaphor, as in (30a), and a free interpretation when it is interpreted from the head position of an NP, as in (30b).

6

This idea is attributable to John Lumsden (research seminar, Fall 1993). Its implementation is mine.

THE RELEXIFICATION ACCOUNT OF CREOLE GENESIS

61

Two independent arguments support this analysis. First, a reciprocal interpretation of the construction may also obtain in both Haitian (e.g. Koopman 1986; Déchaine and Manfredi 1994) and Fongbe (e.g. Kinyalolo 1994), as is shown in (31). (31)

Yo wè yo 0. HAITIAN Yé m yé déè FONGBE they see they SELF 'They saw themselves/each other .' (from Lefebvre 1998a: 167)

The reciprocal interpretation of the Haitian sentence in (31) must result from the fact that Haitian has a covert form corresponding to -déè in Fongbe. Haitian and Fongbe both contrast with French, which requires the clitic se in this context. (32)

a. b.

*Ils voient eux/eux-mêmes FRENCH [Lit.: They see them/themselves.'] (=(69a) in Lefebvre 1998a: 167) Ils se voient. FRENCH They see themselves/each other.' (=(69b) in Lefebvre 1998a: 167)

The second argument supporting the above analysis is that Caribbean creoles whose lexifier language has a SELF anaphor have an overt SELF anaphor. This is the case of English- or Dutch-based creoles. For example, Berbice Dutch has the form -selfu derived from Dutch -zelf (Robertson 1993: 307); Gullah has -self from English -self (Mufwene 1992: 169), and Saramaccan has -seéi (Veenstra 1996a: 43). I now turn to the discussion of BODY-part reflexives as they occur in (25a) and (25b). The phonological representation of Haitian tèt is derived from French tête 'head', and that of ko from French corps 'body'. But in French, these two words cannot be assigned a reflexive interpretation (see Lefebvre 1998a: 167-170). Consequently, the reflexive interpretation of BODY-parts cannot come from French. Are there BODY-part reflexives in Fongbe? Both Kinyalolo (1994) and Brousseau (1995a) report that, in Fongbe, the word meaning 'head' is never assigned a reflexive interpretation, and that, the Fongbe word meaning 'body' wú cannot be assigned a reflexive interpretation either. Consequently, Fongbe cannot be the source of the reflexive interpretation of Haitian ko and tèt. In Lefebvre (1998a: 167-170), it is shown that the reflexive interpretation of BODY-parts may be traced to other substratum languages. More precisely, it is shown that BODY-part reflexives of the type we find in Haitian constitute a widespread phenomenon in Kwa languages (e.g. Awoyale 1986; Faltz 1985; Sylvain 1936). By hypothesis then, speakers of these languages would have used the relexified words for BODY-parts in reflexive constructions.

62

ISSUES IN THE STUDY OF PIDGIN AND CREOLE LANGUAGES

Assuming that this is the correct way of looking at the data, there is no need to appeal to independent development of the Haitian BODY-part forms, as is claimed by Carden and Stewart (1988: 32). The relexification hypothesis provides a straightforward account of the fact that we find several reflexive forms in Haitian. Speakers who had lexicons with reflexive anaphors would use pronominal forms and a phonologically null anaphor when speaking the creole. Speakers who had lexicons with BODY-part reflexives would use BODY-part reflexives when speaking the creole. The variation in the creole reflects differences among the substratum lexicons. This situation suggests that, in the early creole, there were different Haitian dialects reflecting the differences among the substratum languages. The availability of several forms to encode the same notion constitutes an ideal situation for dialect levelling to occur. This topic is taken up in chapter 9. 3.3.1.4 Wh-words. In the languages of the world, Wh-expressions may be generated either by syntactic or morphological rules. In the first case, the Whphrase is headed by a noun that is modified by a Wh-adjective. In the second case, the Wh-phrase is realised as a Wh-word. On the basis of tests distinguishing between Wh-phrases and Wh-words (see Lefebvre 1998a: 171-182), it can be established that Haitian Creole has four Wh-words listed in (33) (phonetic variants are not considered here). Of these four forms, the first two contain the Wh-morphological element ki-. The other two forms are unanalysable simplexes. (33) ki-lès (ki-)sa kouman konbyen

'which one' 'what' 'how' 'how much, how many'

HAITIAN (=(5) in Brousseau 1995a)

As is shown in (33), the form ki-sa 'what' may simply surface as sa as in Sa ou fė? 'What did you do?' (see Valdman et al 1981; Koopman 1982b; Lefebvre 1986; Brousseau 1995a). A specific property of the form ki-lès 'which one' is that it occurs with the plural marker yo when a plural meaning is intended. (34)

Ki-lès yo ou achtel which-one PL you buy 'Which ones did you buy?'

HAITIAN (=(6) in Brousseau 1995a)

The form konbyen 'how much/how many' may occur as the sole element of the Wh-phrase, as in (35a), or it may co-occur with a noun, as in (35b).

THE RELEXIFICATION ACCOUNT OF CREOLE GENESIS (35)

a. b.

63

Konbyen ou achtel HAITIAN how-much/many you buy 'How much/many did you buy?' Konbyen pwason ou achtel HAITIAN how-much/many fish you buy 'How much fish did you buy?' or 'How many fishes did you buy?' (=(7) in Brousseau 1995a)

Questions made out of other positions are encoded by means of syntactic phrases made up of the Wh-adjective ki and a noun, as in (36a). Finally, the Haitian expression meaning 'why' in (36b) is made up of the preposition pou 'for' and the word ki-sa 'what'. In this case, ki-sa may simply be realised as ki as in Pou ki ou fè sa? 'Why did you do that?' (see Lefebvre 1986; Brousseau 1995a). (36)

a.

b.

ki moun 'which person/who' HAITIAN ki bagay 'which thing/what' (ki) kote/ki bo 'which place/where' ki jan 'which manner/how' ki kalite ' which kind/how ' ki lė 'which time, moment/when' pou ki(-sa) 'for what/why' HAITIAN (from Koopman 1982b; Lefebvre 1986; Brousseau 1995a)

Why do Haitian Wh-expressions divide up as they do between Wh-words, as in (33), and syntactically derived Wh-expressions, as in (36)? Why does Haitian have only four Wh-words? And why does it have the particular set it has? Let us first consider the data from the superstratum language. Like Haitian, French has both Wh-words and syntactically derived Wh-phrases. The inventory of French Wh-words used in questions is given in (37). (37)

lequelllaquellellesquelsllesquelles qui quelquoi où quand comment combien pourquoi

'which one(s)' FRENCH 'who' 'what' 'where' 'when' 'how' 'how much/how many' 'why'

Abstracting away from the morphological variants of a single form, the inventory of French Wh-words in (37) comprises eight lexical entries. In contrast, Haitian has only four Wh-words (see (33)). Consequently, in several cases where French has a Wh-word, Haitian has recourse to a syntactically derived Wh-phrase. For example, where French has the lexeme où 'where',

64

ISSUES IN THE STUDY OF PIDGIN AND CREOLE LANGUAGES

Haitian has the syntactic phrase ki kote 'which place'. Like Haitian, French also has Wh-phrases made up of a Wh-adjective and a noun. These are listed in (38a). The Wh-expression in (38b) is made up of the preposition pour 'for' and a Wh-phrase meaning 'which reason'. (38)

a.

b.

quelle personne ?quelle chose (de) quel côté/bord (de) quelle manière quelle sorte (de) quel moment (pour) quelle raison

'which person' ' which thing '7 'which side' 'which manner' 'which kind' 'which time' '(for) which reason'

FRENCH

FRENCH

The structure of the French Wh-phrases in (38a) parallels that of the Haitian Whphrases in (36a). In both languages, the Wh-phrase consists of a Wh-adjective meaning 'which', quel and ki, respectively, and a noun. The structure of the French Wh-expression in (38b), however, is not parallel to that of the Haitian one in (36b). While French has an expression meaning 'for what reason', Haitian has an expression meaning 'for what'. A thorough comparison of the Haitian and French Wh-forms and Wh-expressions in Lefebvre (1998a: 173-182) shows that the details of the Haitian forms in (33) and (36) do not correspond to those of the French ones in (37) and (38). The details of the Haitian forms, however, do follow those of the Fongbe substratum language, to which we now turn. The Wh-words and Wh-phrases of Fongbe (from Anonymous 1983; Brousseau 1995a; Lefebvre 1986 and further data that I collected) are listed in (39). (39)

FONGBE

a. Wh-words:

dè-tέ (é)-tέ/àni nέgbn nàbi

LITERAL GLOSSES

(from Segurola 1963) one-which that-which

MEANING

'which one' 'what' 'how' 'how much/ many'

7 In spoken French, quelle chose is grammatical but it is very unusual to use it in questions of the following type: Quelles choses as-tu achetées? 'What things did you buy?' The complex phrase qu'est-ce que (lit.: 'what is it that') is the expression generally used to question objects.

THE RELEXIFICATION ACCOUNT OF CREOLE GENESIS

b. Wh-phrases:

me

65

tέ tέ

person which 'who' nú thing which 'what' fí(tέ) place which 'where' al tέ manner which 'how' àlkpà tέ kind which 'what kind' hwènù tέ moment/time which 'when' (é)tέ (w)ú(tú)/ what cause 'why' àní(w)ú(tú) (=(92) in Lefebvre 1998a: 177)

Fongbe thus has five Wh-words, two of which are made up of a noun/pronoun and the Wh-affix -tέ. It has two words meaning 'what': (é-)té and àní. Brousseau (1995a) points out that the two forms do not seem to have any distinguishing semantic or syntactic properties except for the fact that àní is less acceptable than (é-)té in echo questions. The other two Wh-words do not contain the Wh-affix -té. The remaining Wh-expressions are syntactic phrases comprised of a noun and the Wh-adjective té 'which'. The Wh-phrase meaning 'why' is made up of the Wh-word meaning 'what' and the postposition (w)ú(tú) 'cause' (see Anonymous 1983). The Fongbe Wh-expressions in (39) are compared with the Haitian ones in (40). (40)

HAITIAN

FONGBE

a. Wh-words: ki-lès dè-té (ki-)sa (é-)télàní kouman négbn konbyen nàbí b. Wh-phrases: ki moun mέ tέ ki bagay nú té (ki) kote/ki bo fi (té) ki jan àlo té ki kalite àlôkpà té ki lè hwènù té pu ki(-sa)

'which one' 'what' 'how' 'how many/much' 'which person/who' 'which thing/what' 'which place/where' 'which manner/how' ' which kind/how ' 'which moment/ time/when' (é)té (w)ú(tú)/ 'what, cause/why' àní(w)ú(tú) (=(93) in Lefebvre 1998a: 178)

Haitian and Fongbe have inventories of only four and five Wh-words, respectively. Except for àni 'what', which has no corresponding form in

66

ISSUES IN THE STUDY OF PIDGIN AND CREOLE LANGUAGES

Haitian, each of the Fongbe forms has a Haitian equivalent. Furthermore, the morphological makeup of these forms is similar in the two languages. The Fongbe forms which include the Wh-affix -té correspond to the Haitian forms with the Wh-affix ki-. In both languages, the other two Wh-words do not contain this Wh-morpheme. Like its Haitian counterpart, the Fongbe form dè-té 'which one' is unmarked for gender and requires the plural marker le when a plural meaning is intended. The Fongbe data in (41) correspond to the Haitian data in (34). (41)

Đè-té lέ à which one PL you buy 'Which ones did you buy?'

xo?

FONGBE (=(11) in Brousseau 1995a)

Like the Haitian form konbyen 'how much/many', the Fongbe form nàbi 'how much/many' may be the sole lexical element in its projection, or it may occur with a noun, as in (42), which parallels the Haitian data in (35). (42)

a.

b.

Nàbí à xol how-much/many you buy 'How much/many did you buy?' Hwèví nàbí à fish how-much/many you buy 'How many fish did you buy?' or 'How much fish did you buy?'

FONGBE

x

FONGBE

(=(12) in Brousseau 1995a)

In both languages, all the other Wh-expressions are syntactic expressions involving the Wh-adjectives ki and té, respectively, plus a noun. Furthermore, in both languages, the Wh-expressions have the same meaning. Finally, the forms (é)tέ (w)ú(tú) and pou ki-(sa) 'why' parallel each other in an interesting way: both involve a Wh-word, meaning 'what', and a lexical element of the category P: the postposition (w)ú(tú) in Fongbe, meaning 'cause', and the preposition pou in Haitian meaning 'for'. Thus, the reason why the Haitian Wh-expressions divide up as they do, between Wh-words and Wh-phrases, follows from the relexification hypothesis of creole genesis. Haitian has Wh-words which correspond to lexical entries in the substratum lexicon. This supports the argument that the creators of Haitian relabelled their own lexical entries using French phonetic matrices, and that they ignored the French forms that they did not have in their own lexicon. This explains why the French simplexes qui 'who', que 'what', où 'where', quand 'when' and pourquoi 'why' did not make their way into Haitian. The relexification hypothesis also explains why the morphological makeup of the

THE RELEXIFICATION ACCOUNT OF CREOLE GENESIS

67

Haitian Wh-words is so similar to that of the corresponding Fongbe words (see (40)). Finally, Haitian has Wh-phrases exactly where the substratum language has Wh-phrases, showing that the creators of Haitian used the concatenating properties of their own grammar and lexicon in creating the Haitian Creole Whphrases. In Lefebvre (1998a: 171-182), it is further argued that the semantic, syntactic and distributional properties of the Wh-expressions of Haitian Creole follow the details of its substratum languages. One exception to this general state of affairs is discussed: the relative position of the Wh-element and the head of the construction in which it appears. While ki precedes the head of the construction in Haitian, té follows it in Fongbe. This discrepancy, however, is exactly what is expected under the theory of how word order is established in creole genesis (see section 3.1.4). For lexical categories, the word order of the creole is predicted to follow that of the superstratum language. In French, the Wh-adjective quel 'which' precedes the noun. The position of Haitian ki follows the French pattern. Likewise, while the Fongbe Wh-phrase meaning 'why' makes use of the postposition (w)ú(tú), the corresponding Haitian expression makes use of a preposition, in accordance with the French word order. The three-way comparison presented above shows that the semantic and syntactic properties of the Haitian Wh-expressions are derived from those of the corresponding lexical entries in the substratum language. Their phonological representations are derived from phonetic strings found in the superstratum language. This division of properties is exactly what is predicted by the relexification hypothesis. 3.3.1.5 Verbs. As is the case of the nouns in (6), the Haitian verbs in (43) have two meanings. Their phonological representations are derived from corresponding French verbs. These French verbs, however, lack one of the two meanings associated with the Haitian verbs. This shows that French cannot be the source of the extra meaning associated with the Haitian verbs. Fongbe verbs, however, appear to be the source of the extra meaning associated with the Haitian verbs. Again, this follows from the relexification hypothesis. (43)

HAITIAN

FRENCH

ansasinen 'to murder' 'to mutilate' gade 'to concern'

assassiner 'to murder' regarder 'to concern' 'to look'

FONGBE

hù 'to murder' 'to mutilate' kán 'to concern'

68

ISSUES IN THE STUDY OF PIDGIN AND CREOLE LANGUAGES

gode 'to watch over/ to take care of' 'to keep' 'to look' 'to imitate' kase 'to slim down' 'to break' kraze 'to break to pieces' 'to break by spreading' 'to disperse'

garder kpn 'to watch over/ 'to watch over/' to take care of' to take care of' 'to keep' 'to keep' regarder 'to look' 'to look' imiter 'to imitate' 'to imitate' casser gbà (kpô) 'to slim down' 'to break' 'to break' écraser Jajá 'to destroy' 'to break to pieces' 'to crush' 'to break by spreading' 'to disperse' (from Lefebvre 1999a: 69-79)

Many other examples of this type illustrating cases of substratum semantics in Haitian Creole resulting from relexification can be found in Brousseau (1989), Lefebvre (1998a, 1999a), Lumsden (1999a, 1999b). Furthermore, in Lefebvre (1999a, and the references therein), it is shown that, where the aspectual properties of verbs are not the same in the three languages under comparison, Haitian generally pairs with Fongbe rather than with French. Likewise, thematic properties of Haitian verbs (with respect to agentiveness) are shown to follow those of Fongbe rather than those of French, in cases where Fongbe and French differ. In spite of these extensive similarities between Haitian and Fongbe, there are also some differences between the two languages. For example, as is discussed in Lefebvre (1999, and the references cited therein), there are more verbs in Haitian than there are in Fongbe. This situation is hypothesised to result from the acquisition, by the creators of Haitian, of French verbs that have no counterpart in the Haitian substratum languages. The consequences of this state of affairs on the makeup of the creole is a topic for future research. 3.3.1.6 Summary. The data presented in this section illustrate various effects of the role of the process of relexification in creole genesis. First, they illustrate the systematic division of properties of the creole lexical entries between its contributing languages: while the label of the Creole's lexical entries are derived from its superstratum language, its semantic properties are derived from the substratum languages. Second, they illustrate the fact that some French lexical entries did not make it into the creole because the creators of the creole did not have corresponding lexical entries to relexify (e.g. the second person plural personal pronoun, the possessive adjectives and pronouns, etc.). Third, they show that lexical entries that do not have independent semantic content cannot be

THE RELEXIFICATION ACCOUNT OF CREOLE GENESIS

69

relabelled (e.g. the logophoric pronouns of the Haitian substratum languages). Fourth, they show that the concatenation of words into compounds follows the principles of the substratum languages rather than those of the superstratum. Given this conclusion, one might wonder why Haitian Creole appears to make fewer lexical distinctions, in some areas of the lexicon, than its substratum languages. For example, as is shown in Lefebvre (1989), while Fongbe has several verbs of cutting encoding various manners or cutting, Haitian has fewer verbs participating in the paradigm of cutting verbs. In Lefebvre (1989) and in Lefebvre (1998a), it is extensively shown that, as a result of the discrepancy between the substratum and the superstratum lexicons, several substratum lexical entries may end up being relabelled by the same superstratum form. On the surface, such cases may appear as counterexamples to the relexification hypothesis of creole genesis. Through deeper analysis, however, such cases illustrate the constraints imposed by the superstratum language on the process of relabelling. 3.3.2 The syntactic properties of verbs Although the semantic and syntactic properties of verbs are not always easy to distinguish from each other, there are properties of verbs that are generally considered to be syntactic, such as selectional, raising, control, and caseassigning properties. Verbs' syntactic properties are specified in their lexical entries. The relexification hypothesis predicts that the syntactic properties of verbal lexical entries in the substratum languages will be reproduced in the corresponding lexical entries in the creole. The comparison of the syntactic properties of Haitian, French and Fongbe verbs shows that, to a great extent, this prediction is borne out. 3.3.2.1 Types of argument structures. Detailed discussions of types of argument structures in Haitian and Fongbe can be found in Massam (1989) and Lefebvre (1991a), respectively. A preliminary comparison of argument structures in Haitian, French and West African languages may be found in Koopman (1986). Building on these findings and on further work, Lefebvre (1998a: 248-250) provides a short overview of the types of argument structures in Haitian, French and Fongbe, pointing out their similarities and differences. All three languages have monadic verbs. These include unaccusative verbs such as 'to go', 'to come', etc. While such Haitian and Fongbe verbs occur in their bare form, as in (44a), the corresponding French verbs are conjugated with the auxiliary 'to be', as in (44b).

70 (44)

ISSUES IN THE STUDY OF PIDGIN AND CREOLE LANGUAGES a. b.

Li ale / rive. É yi / wá. 'He left/arrived.' Il est parti I arrivé. 'He left/arrived. '

HAITIAN FONGBE FRENCH (=(1) in Lefebvre 1998a: 249)

In both Haitian and Fongbe the locative argument of these verbs need not be introduced by a case marker or a pre- or postposition, whereas in French, the locative object has to be introduced by à 'at'. This contrast is shown in (45) a and b, respectively. (45)

a. b.

Li ale / rive Pdtoprens. HAITIAN É yi / wá Kùtnû. FONGBE 'He went to/arrived in Port au Prince/Cotonou.' Il est allé I arrivé à Paris. FRENCH 'He went to/arrived in Paris.' (=(2) in Lefebvre 1998a: 249)

Single-variable verbs also include unergative verbs such as 'to bark' and 'to jump', as shown in (46). While such Haitian and Fongbe verbs occur in their bare form, the French equivalents are conjugated with the auxiliary 'to have'. (46)

a.

b.

Li abwaye / sote. É hó / In. 'He barked/jumped.' Il a aboyé / sauté. 'He barked/jumped.'

HAITIAN FONGBE FRENCH (=(3) in Lefebvre 1998a: 249)

As is pointed out in Lefebvre (1998a: 249), the range of single-variable verbs is smaller in Fongbe than in French and Haitian. This is because a concept rendered by a single-variable verb in Haitian and French is sometimes expressed by a light verb construction or an inherent object verb in Fongbe (see below). Single-variable predicates also include WEATHER verbs, some verbs selecting expletive subjects, some raising verbs, and existential verbs.These will be discussed below. The second type of argument structure consists in two-variable (or transitive) verbs as in He ate bread. All three languages have a large class of these verbs. Finally, all three languages have three-variable (or ditransitive) verbs. These include predicates of transfer such as 'to give'. In Haitian and Fongbe, these predicates are rendered by a double-object construction, as in John gave Mary a book, or a serial verb construction. As will be shown below, French does not have such constructions and three-variable predicates are rendered by a

THE RELEXIFICATION ACCOUNT OF CREOLE GENESIS

71

construction of the type John gave a book to Mary. Three-variable predicates also include some control verbs, discussed below. It will be shown that the properties of Haitian and Fongbe verbs are quite similar and contrast with those of the corresponding French verbs. 3.3.2.2 BODY-state verbs. In Haitian, BODY-state expressions such as I have a headache are built on the model BODY-part + VERB + PRONOUN, as shown in the following examples. (47) a.

b. c.

Vont mwen ap fè m mal. HAITIAN stomach me IMP do me hurt 'I have a stomachache.' [Lit.: 'My stomach hurts me.'] (=(29) in Koopman 1986) Tèt mwen ap fè m mal. HAITIAN head me IMP do me hurt 'I have a headache.' (=(29) in Koopman 1986) Dan ap manje m. HAITIAN tooth at eat me 'I have a toothache.' [Lit.: 'My tooth is eating me.'] (=(4c) in Lefebvre 1998a: 250)

French also allows BODY-state expressions on the model of the Haitian ones in (47). For example, it is possible to say L'estomac me brûle (lit.: 'The stomach burns me'), La tête me fait mal (lit.: 'The head hurts me'). However, BODY-state expressions are typically built on the model X HAVE PAIN at BODY-PART. (48)

a. b. c.

J' ai mal à la tête. I have pain at DEF head 'I have a headache.' J' ai mal au ventre. I have pain at belly T have a stomachache.' J' ai mal aux dents. I have pain at teeth T have a toothache.'

FRENCH FRENCH FRENCH (=(5) in Lefebvre 1998a: 250)

The French expressions in (48) have no counterpart in Haitian. As is pointed out by Koopman (1986), the structure of the Haitian expressions in (47) is similar to corresponding expressions in the West African languages, which are also built on the model BODY-part + VERB + PRONOUN. Examples from Fongbe are provided in (49).

72 (49)

ISSUES IN THE STUDY OF PIDGIN AND CREOLE LANGUAGES a. b. c.

Xómè wli mi. FONGBE stomach hold me 'I have a stomachache.' [Lit.: 'My stomach is holding me.'] Ta dù mi. FONGBE head eat me 'I have a headache.' [Lit.: 'My head is eating me.'] ÀDÚ DÒ dúdú mi wέ. FONGBE tooth be.at eat me POST 'I have a toothache.' [Lit.: 'My tooth is eating me.'] (=(6) in Lefebvre 1998a: 251)

BODY-state verbs in Haitian (see (47)) and Fongbe (see (49)) typically select BODY-parts as their subject. This contrasts with French BODY-state verbs, which typically take a pronominal subject (see (48)). This should come as no surprise given the relexification hypothesis. 3.3.2.3 WEATHER verbs. As has been pointed out by Koopman (1986) and documented in detail by Dumais (1988), Haitian expresses various atmospheric phenomena by means of a construction that uses verbs (which occur in other contexts as well) selecting a lexical subject referring to a natural element. This is shown in (50). (50) a. Lapli tonbe. rain fall 'It is raining.' [Lit: 'Rain falls.'] b. Laglas tonbe. ice fall 'It is hailing.' [Lit.: 'Ice falls.'] c. Yon ti- van vante. a little wind wind 'It is windy.' [Lit.: 'The wind winds.'] d. Ldraj gwonde. storm growls 'It is thundering.' [Lit.: 'The storm growls.'] e. Lapli op farinen. rain IMP drizzle 'It is drizzling.' [Lit.: 'Rain is drizzling.']

HAITIAN (=(1) in Dumais 1988) HAITIAN (=(2) in Dumais 1988) HAITIAN (=(3) in Dumais 1988) HAITIAN (=(4) in Dumais 1988) HAITIAN (from Valdman et al. 1981)

Both authors point out that the above Haitan data contrast with French, where the same concepts are rendered by means of WEATHER verbs selecting an

THE RELEXIFICATION ACCOUNT OF CREOLE GENESIS

73

expletive subject, as shown in (51). An expletive subject is not allowed in Haitian in the context of the WEATHER verbs in (50). (51)

a. Il pleut 'It is raining. ' c. Il vente. 'It is windy.' e. Il bruine. 'It is drizzling.'

b. Il grêle. 'It is hailing. ' d. Il fait un orage./Il tonne. 'It is stormy./It is thundering.'

FRENCH FRENCH FRENCH

As is pointed out in Lefebvre (1998a: 252), some French expressions built on the model of the Haitian ones in (50) may be grammatical; for example, one may find La pluie tombe 'Rain is falling', L'orage gronde 'The storm is growling' but not *Un petit vent vente 'A little wind is blowing' nor *La bruine bruine 'Drizzle is drizzling.' As the translations show, however, these expressions can only be interpreted literally. Furthermore, the first two expressions can only be used in specific contexts for stylistic effects (e.g. in poetry). The standard way of using WEATHER verbs in Haitian (see (50)) thus differs from the standard way of using WEATHER verbs in French (see (51)). Again, the properties of the Haitian WEATHER expressions will be shown to follow the pattern of the substratum languages. Koopman (1986: 245) points out that WEATHER verbs selecting an expletive subject do not exist in West African languages any more than they do in Haitian. On the basis of examples from Vata and Abe, she shows that West African languages generally express the various atmospheric phenomena in a construction involving verbs (that occur in other contexts as well) selecting a lexical subject referring to a natural element. The Fongbe data in (52) illustrate this pattern. (52)

a. b. c. d.

Ji jà. rain falls 'It is raining.' [Lit.: 'Rain falls.'] Láglási jà. ice fall 'It is hailing.' [Lit.: 'Ice falls.'] Jòhn nyi. wind blow 'It is windy.' Hèbyôsô dó-gbè. thunder growl 'It is thundering.'

FONGBE (=(1) in Dumais 1988) FONGBE (=(2) in Dumais 1988) FONGBE (=(3) in Dumais 1988) FONGBE (=(9d) in Lefebvre 1998a: 252)

74

ISSUES IN THE STUDY OF PIDGIN AND CREOLE LANGUAGES e.

Ji

rain drizzle 'It is drizzling.'

dè-àtân.

FONGBE

(=(9e) in Lefebvre 1998a: 252)

The Haitian expressions in (50) are built on the model of expressions in the West African languages like those in (52). The data in (50) and (52) reflect the selectional properties of the verbs involved. In both Haitian and Fongbe, but not in French, verbs meaning 'to fall', 'to be windy', 'to thunder', 'to drizzle' take an argument which is a natural element such as 'rain'/'ice'/'wind', etc. The Haitian facts are predicted by the relexification hypothesis. (For further discussion, see Lefebvre 1998a: 253.) 3.3.2.4 Reflexive verbs. Recall from section 3.3.1.3 that, in Haitian, a reflexive interpretation may be induced by a pronoun + SELF (where SELF is phonologically null), tėt- + PRONOUN (lit.: 'X's head'), or ko- + PRONOUN (lit.: 'X's body'). A few verbs also allow for a reflexive interpretation (without any overt reflexive form) when their internal argument is not realised in the syntax (e.g. abiye 'to dress oneself). Recall also that the various ways of encoding reflexivity in Haitian come from various substratum languages. This suggests that more than one reflexive form may be selected by a given verb in Haitian. As is shown in the literature on Haitian, this is the case (see Carden and Stewart 1988; Faine 1937; Goodman 1964; Sylvain 1936). As is shown in Brousseau (1995b), there are even verbs (e.g. blese 'to hurt') which may select all four forms. The same situation is also observed in other Caribbean creoles (see Muysken and Smith 1995). Lefebvre (1998a: 255) points out that this is probably due to the fact that, as we saw in section 3.3.1.3, the different reflexive forms found in a given creole have been transferred into it from a variety of substratum languages through relexification. Since each individual substratum language has a different subset of the total inventory of forms found in the substratum languages as a group, the prediction is that the subcategorisation properties of verbs (considered as a whole) for reflexive forms in a given creole should not necessarily match those of the corresponding verbs in any of the substratum languages taken individually. Brousseau's (1995b) comparison of the subcategorisation properties of verbs for reflexive forms in Haitian and Fongbe shows that this prediction is borne out. Lefebvre (1998a: 253-262) extensively argues that this situation follows from the relexification hypothesis and subsequent levelling. 3.3.2.5 Verbs licensing expletive subjects. As we saw in section 3.3.1.2.4, all three languages under comparison allow for expletive subjects. In Lefebvre (1998a: 259-260), it is shown that the verbs allowing for expletive subjects are

THE RELEXIFICATION ACCOUNT OF CREOLE GENESIS

75

the same in all three languages (i.e. 'to seem', 'to remain', 'to be missing', 'to be good'). In addition to having an overt expletive, Haitian and Fongbe, but not French, have a covert one, as we saw in section 3.3.1.2.4. Are the selectional properties of verbs the same in Haitian and Fongbe with respect to overt/covert expletives? Lefebvre (1998a: 261-262) shows that only two out of seven pairs of Haitian/Fongbe verbs have the same selectional properties. The other pairs show various type mismatches, and more Haitian verbs allow for a null expletive. Considering that the covert expletive option is a property of Haitian inherited from the substratum languages, the latter observation may be considered surprising. It is suggested that there is probably variation across Haitian speakers and across West African languages with respect to which kind of expletive subjects (overt or covert) verbs select. This is an area of the lexicon where dialect levelling is likely to occur. As is pointed out in Lefebvre (1998a: 262), further documentation of the pertinent facts needs to be done. 3.3.2.6 Raising verbs. Verbs which licence expletive subjects also allow for argument raising to subject position. Argument raising may proceed from an object position, as in An apple remains in the basket (
a. b.

Li rete yon sėpan {nan pannye an). HAITIAN it remain a snake in basket DEF 'There remains a snake (in the basket).' (=(14) in Dumais 1988) Yon sėpan rete nan pannye an. HAITIAN a snake be.left.over in basket DEF 'A snake remains in the basket.' (=(15) in Dumais 1988)

Dumais (1988) points out that the Haitian data in (53) contrast with the French data involving the verb rester 'to remain'. As is shown in (54a), the argument of the French verb rester appears in the position following the verb and the subject position is obligatorily filled by the expletive form iL The ungrammaticality of the (b) sentence shows that the argument of rester cannot appear in the subject position when the verb establishes a locative relation.

76 (54)

ISSUES IN THE STUDY OF PIDGIN AND CREOLE LANGUAGES a. b.

Il

reste un serpent dans le panier. FRENCH it remain a snake in DEF basket There remains a snake in the basket.' *Un serpent reste dans le panier FRENCH a snake remain in DEF basket (=(30) in Lefebvre 1998a: 263)

The ungrammaticality of the French sentence (54b) contrasts with the grammaticality of the Haitian sentence (53b). Thus, although the French verb rester is the phonetic source of rete, it did not contribute its syntactic properties. These properties, however, may be argued to come from the substratum languages. For example, in Fongbe, there is a verb kpò, which means both 'to remain' and 'to be left over' (Segurola 1963). As is the case in Haitian, the argument of this verb may occupy two surface positions. In (55a), it occurs after the verb and the subject position is filled with the pleonastic form é. In (55b), it occurs in the verb's subject position. (55)

a. b.

É kpò dàn dókpó (dò xàsùn  me). FONGBE it remain snake one (be.at basket DEF in) 'There remains a snake in the basket.' (=(14) in Dumais 1988) Dàn dókpó kpô (dò xàsùn  mè). FONGBE snake one remain (be.at basket DEF in) (=(15) in Dumais 1988) 'A snake remains in the basket.'

As Dumais (1988) shows, in Fongbe, the locative phrase is optional regardless of the surface position of the argument. The optionality of the locative phrase in (55b) contrasts with its obligatoriness in the Haitian example in (53b). Except for this difference, however, both Haitian rete and Fongbe kpô allow the argument to occupy two surface positions. It thus appears that the syntactic properties of the Haitian verb rete follow those of the corresponding verb kpô in the substratum language rather than those of the French verb rester which is its phonetic source. Similar data involving the verbs meaning 'to be missing' in the three languages under comparison present a similar pattern: while the Haitian and Fongbe verbs allow object to subject raising, the French verb does not (see Lefebvre 1998a: 262-269). As for subject to subject raising, French is like English in allowing it only out of an infinitival clause. As is extensively discussed in Lefebvre (1988a: 266-269), based on Dumais (1988), Massam (1989) and Law (1992), Haitian is like Fongbe, allowing subject raising out of tensed clauses. Massam (1989) observes that the type of subject raising manifested in Haitian (and in Fongbe) is extremely rare. This makes it a marked phenomenon. The fact that parallel data exist in Fongbe shows that a marked feature of the substratum language has been

THE RELEXrFICATION ACCOUNT OF CREOLE GENESIS

77

transferred into the creole. On the assumption that subject raising out of an infinitival clause is the unmarked option, we are left with the fact that Haitian has retained the marked option from the substratum languages instead of adopting the unmarked option from the superstratum language. The syntactic properties of the Haitian raising verbs thus appear to follow in a straightforward way from the relexification hypothesis. 3.3.2.7 Existential verbs. A detailed comparison of the three existential verbs in the languages under comparison in Lefebvre (1998a: 269-271) shows that the Haitian existential verb gen has properties that differ from the corresponding verbs in both of its source languages. Given that the properties of gen resemble neither those of French nor those of Fongbe, we may ask whether they could be derived from an equivalent lexical entry in some other West African language. Koopman (1986: 248) points out, however, that no West African language has a form with the properties of Haitian gen. Following the methodology adopted for the research, we therefore have to conclude that gen is an innovation. This conclusion is in agreement with that in DeGraff (1992b). 3.3.2.8 Control verbs. Control verbs are verbs which allow an argument of a matrix sentence to be coindexed with an empty argument position in the complement clause of the matrix verb. The three-way comparison of Haitian, French and Fongbe shows striking similarities between Haitian and Fongbe, both of which contrast with French in the same way. Two-variable Haitian control verbs of the WANT-class may select either a tensed or an infinitival complement. For example, the tensed complement of vle 'to want' is introduced by the complementiser pou (see section 3.3.5.2). The embedded subject may be either coreferential with the matrix subject or disjoint in reference from it (see Koopman 1986; Sterlin 1988, 1989). This is shown in (56). (56)

Lii vle pou li/j vini. HAITIAN he want COMP he come 'He wants to come.' or 'He wants him to come.' (=(18b) in Koopman 1986)

Koopman (1986: 240) points out that only one of the two interpretations available in Haitian is available in French. In contrast to Haitian, the tensed complement of vouloir 'to want' only allows for disjoint reference of the subject, as shown in (57).

78

(57)

ISSUES IN THE STUDY OF PIDGIN AND CREOLE LANGUAGES

Ili veut qu ilj vienne. FRENCH 3rd want COMP 3rd come 'He wants him to come.'/#'He wants to come.' (=(15b) in Koopman 1986)

The Haitian data in (56) are, however, parallel to the Fongbe data. The Fongbe verb jló 'to want', in (58), selects a tensed complement introduced by the form nú, which corresponds to Haitian pou (see section 3.3.5.2). As is the case in Haitian, the embedded subject can be either coreferential with the matrix subject or disjoint in reference from it. (58)

Éi jló nú éi/j ni yî. FONGBE 3rd want COMP 3rd SUB leave 'He wants to leave.' or 'He wants him to leave.' (=(49) in Lefebvre 1998a: 272)

Koopman (1986: 241) provides similar examples from Vata and Akan, other substratum languages of Haitian. Thus, the Haitian verbs of the WANT-class do not have the syntactic properties of the French verbs that they were phonologically derived from. They do, however, have the same properties as the Fongbe verbs that they were relexified from. In all three languages under comparison, verbs of the WANT-class may also take an infinitival complement in which the matrix subject binds a position in the embedded clause (for Haitian, see Koopman 1986; Sterlin 1988, 1989; for Fongbe, see Kinyalolo 1992, Lefebvre 1993a). In the examples in (59), the controlled position is represented as PRO, a label which stands for an abstract pronoun that is not pronounced but that has referential properties. (59)

a.

b.

c.

Jani vle PROi kraze manchin-nan. HAITIAN John want destroy car DEF 'John wants to destroy the car.' (=(50a) in Lefebvre 1998a: 272) Jeani veut PROi détruire V auto. FRENCH John want destroy DEF car 'John wants to destroy the car.' (=(50b) in Lefebvre 1998a: 272) Kkúi jló PROi na gbà mto . FONGBE Koku want DEF.FUT destroy car DEF 'Koku wants to destroy the car.' (=(50c) in Lefebvre 1998a: 272)

The most interesting facts about the infinitival complements of verbs of the WANT-class in the three languages under comparison are shown in (60). In

THE RELEXEFICATION ACCOUNT OF CREOLE GENESIS

79

addition to taking an infinitival complement of the type in (59), both Haitian and Fongbe verbs, but not the French verb, may select an infinitival complement containing an overt subject. Unlike the embedded covert subject in (59), however, the embedded overt subject in (60) must be referentially disjoint from the matrix subject. (60)

a. b. c.

Lii vle li vini. 'He wants him to come.' *Il veut lui venir [Lit.: 'He wants him to come.'] Éi bà èj yï. 'He wants him to come.'

HAITIAN (=(8) in Sterlin 1988) FRENCH (=(5 lb) in Lefebvre 1998a: 273) FONGBE (=(51c) in Lefebvre 1998a: 273)

In Lefebvre (1998a: 272-276, and the references therein), it is extensively argued that the Haitian and Fongbe data in (60) can be accounted for in a unified way: the subject of the embedded clause is assigned accusative case by the matrix verb under Exceptional Case Marking. Thus, while French verbs of the WANT-class can select only one type of infinitival complement (without an overt subject), both Haitian and Fongbe select two types of such complements. In the first type, there is no overt subject but the subject of the embedded verb is understood as being coreferential with that of the matrix clause (see (59)). In the second type, there is an overt subject and, in both languages, this subject must be referentially disjoint from the matrix subject (see (60)). The selectional properties of Haitian verbs must follow from the relexification hypothesis. Furthermore, the availability of an overt subject is attributable to the fact that the Haitian and Fongbe verbs can assign accusative case to the subject of the embedded clause, a property that French verbs do not have. This syntactic property of Haitian verbs must follow from the relexification hypothesis since it is a property of substratum but not superstratum verbs. As for the selectional properties of three-variable control verbs, such as 'to promise', 'to ask/request', in Haitian, French and West African languages, Koopman (1986: 242) remarks: "The picture that emerges is clear: although the phonetic shape of the Haitian verbs is clearly derived from French, their selectional properties are rather different from those of French, and strikingly similar to those observed in West African languages." Data illustrating this state of affair are provided in Lefebvre (1998a: 276-278). The data show that three-place control verbs in Haitian and Fongbe share the same selectional properties and contrast with the corresponding French verbs. This follows straightforwardly from the relexification hypothesis. Koopman (1986: 240) further remarks that three-place control verbs

80

ISSUES IN THE STUDY OF PIDGIN AND CREOLE LANGUAGES

corresponding to French verbs such as convaincre 'to convince', ordonner 'to order', and persuader 'to persuade' are difficult or impossible to find in both Haitian and the West African languages. This gap in the Haitian lexicon also follows from the relexification hypothesis: the creators of Haitian simply did not have such lexical entries to relexify. While the selectional properties of Haitian control verbs differ from those of the corresponding French verbs, they are the same as those of the Fongbe verbs. In both Haitian and Fongbe, but not in French, the subject of the tensed complement of verbs of the WANT-class may be coreferential with the matrix subject. In both Haitian and Fongbe, but not in French, verbs of the WANT-class may select an infinitival complement with an overt subject which is assigned accusative case by the main verb under Exceptional Case Marking. Finally, three-variable control verbs in Haitian and Fongbe share properties which distinguish them from French. The syntactic properties of Haitian verbs thus must have been transferred into the creole through relexification. 3.3.2.9 Light verbs. Light verb constructions involve a verb and an object as in take a walk in English. A major characteristic of these constructions is that the verb contributes very little to the semantics of the construction; rather, it is the object that is determinative. All three languages under comparison have light verb constructions, as shown in (61). (61)

FONGBE

FRENCH

HAITIAN

fun àhwàn

faire la guerre

ß

lage

'make war'

In Fongbe, however, there are a number of light verb constructions which correspond to Simplexes in French and Haitian. Examples are provided in (62). The Fongbe data are from Brousseau (1988), the Haitian data from my own field notes. (62)

FONGBE

FRENCH

HAITIAN

wà àz 'do work' kú t 'die river' je tàgbà 'fall problem' je àkpà 'fall wound' dó gàn 'plant metal'

travailler

travaye

'to work'

se noyer

nwaye

'to drown'

s'inquiéter

enkyete

'to worry'

se blesser

blèse

'to hurt oneself'

enchaîner

anchènnen

'to chain up'

THE RELEXIFICATION ACCOUNT OF CREOLE GENESIS

fi hàn chanter 'give birth song' su xò crier 'close speech' blá nù jeûner 'tie mouth' kán wèzùn courir 'pick up run'

chante

'to sing'

kriye

'to shout'

jennen

'to fast'

kouri

81

'to run' (=(62) in Lefebvre 1998a: 279)

The question here is whether the data in (62) constitute evidence for or against the relexification hypothesis. The answer to this question rests on whether light verbs and their objects constitute lexical entries or not. In the recent literature, light verbs have been analysed as complex predicates listed in the lexicon (e.g. Cattell 1984; Grimshaw and Meister 1988; Travis in press). On the basis of phonological and syntactic arguments (e.g. various types of extraction facts), Brousseau (1988) extensively argues that the objects of the light verbs in (62) differ from ordinary objects. She concludes that Fongbe light verb constructions must therefore be listed in the lexicon. Since they are listed in the lexicon, they should undergo relexification. How does relexification proceed in this case? Substratum speakers who had lexical entries like the Fongbe ones in (62) searched in the superstratum language for phonetic strings to relabel the lexical entries copied from their own lexicon. They found the French Simplexes in (62) and used them for this purpose, yielding the Haitian lexical entries in (62). The fact that simplexes in the superstratum language may be used to relabel light verb constructions during relexification is in line with the analysis that these constructions constitute lexical entries. In turn, this is additional evidence for analyses holding that light verb constructions are complex predicates that are listed in the lexicon. This conclusion is further reinforced by the fact that, when both the substratum and superstratum languages encode a notion with a light verb construction, the creole ends up with a light verb construction as well, as shown in (61). 3.3.2.10 Inherent object verbs. Inherent object verbs are verbs which are semantically autonomous. Unlike the objects of light verbs, their objects do not contribute to the meaning of the verb itself. These verbs can take different types of objects, but they cannot surface without an overt object of some kind. When no specific object is intended, these verbs will appear with the typical object that is appropriate for a given verb, or an object meaning 'thing'. Fongbe has several such verbs. Their objects may be cognate to the verb, as in no and (lit.: 'suck breast') 'to suckle', or not, as in kùn hun (lit.: 'drive vehicle') 'to drive'. On the basis of data from English, Massam (1990) argues that cognate objects behave

82

ISSUES IN THE STUDY OF PIDGIN AND CREOLE LANGUAGES

like ordinary objects and consequently cognate object verbs are listed in the lexicon independently of their object. Based on Fongbe data, Brousseau (1988) argues that the objects of inherent object verbs share syntactic characteristics with ordinary direct objects, whether or not they are cognate to the verb. Her conclusion is thus the same as Massam's. These analyses predict that, in relexification, inherent object verbs will be relexified independently of their objects. Since the property of taking an inherent object is a property of verbs, we expect the Haitian verbs corresponding to inherent object verbs in the substratum to reproduce this property. Is this prediction borne out by the data? In Lefebvre (1998a: 280), it is shown that this prediction is only partially born out by the data, as only three Haitian verbs follow the pattern of the substratum language, requiring an inherent object (see (63a)). In (63b) the inherent objects of the substratum verbs are not reproduced in the creole any more than the cognate objects in (63c). (63)

FONGBE

a. gb àzn 'calm disease' nyà gbé 'hunt animals' dùn sin 'draw water' b.kùn hún 'drive vehicle' DÙ nú 'eat thing' zà àyi 'sweep ground' c. n an 'suck breast' d ad 'pee piss' kpen àkpen 'cough cough' di àdi 'believe belief' kwin àkwin 'whistle whistle' dù àdù 'eat food'

FRENCH

HAITIAN

calmer douleur

kalme doulè

'to heal'

chasser

chase bet

'to hunt'

puiser conduire

tire dlo 'to draw 'draw water' (water)' kondwi 'to drive'

manger

manje

'to eat'

balayer

bale

'to sweep'

téter

tete

'to suckle'

pisser

pise

'topee'

tousser

touse

'to cough'

croire

kwè

'to believe'

siffler

sifle

'to whistle'

gagner

genyen 'to win' (=(63) in Lefebvre 1998a: 280-281)

THE RELEXIFICATION ACCOUNT OF CREOLE GENESIS

83

In Lefebvre (1998a: 281), it is proposed that the difference between the Fongbe verbs in (63 a and b) and their corresponding verbs in French and in Haitian may best be stated in terms of their transitivity properties: whereas the Fongbe verbs are necessarily transitive, the Haitian and French verbs may also be used intransitively. Lefebvre (1998a: 282-283) provides a theoretical account of discrepancy between Fongbe, on the one hand, and Haitian and French, on the other hand. She concludes that, with regard to this property, Haitian follows the option of the superstratum language rather than that of the substratum language. 3.3.2.11 The case-assigning properties of verbs. A comparison of the case properties of some hundred triplets of verbs in Lefebvre (1998a: 283-287) shows that the case-assigning properties of Haitian verbs are quite free as compared with the corresponding verbs in both its superstratum and substratum languages. A discussion of why this should be the case can be found in Lefebvre (1998a: 286). 3.3.2.12 Double-object verbs. John sent Mary a letter is an example of the recipient-theme construction (NP NP), whereas John sent a letter to Mary is an example of the theme-locative construction (NP PP). Like West African languages, Haitian Creole has the recipient-theme construction (e.g. Koopman 1986; Lumsden 1994; Veenstra 1992). It is a well-documented fact, however, that French does not (e.g. Kayne 1984; Tremblay 1991). Fongbe, however, does have the recipient-theme construction (e.g. Lefebvre 1994c). The contrast between the three languages is illustrated in (64), (65) and (66). (64) Mwen bay I montre Pol liv la. HAITIAN I give/ show Paul book the T gave/showed Paul the book.' (=(19) and (25b) in Lumsden 1994) (65)

a.

b.

*J' ai donné I montré Paul le livre FRENCH I AUX give / show Paul DEF book [Lit.: T gave/showed Paul the book.'] (=(81) in Lefebvre 1998a: 290) J' ai donné I montré le livre à Paul. FRENCH I AUX give / show DEF book to Paul T gave/showed the book to Paul.' (=(82) in Lefebvre 1998a: 290)

(66) Koku na / xélé Àsíbá xwé 6. Koku give / show Asiba house DEF 'Koku gave/showed Asiba the house.'

FONGBE (=(la, b) in Lefebvre 1994c)

84

ISSUES IN THE STUDY OF PIDGIN AND CREOLE LANGUAGES

These constructions are extensively discussed in Lefebvre (1998a: 287-301). In addition to addressing tests distinguishing between the NP NP and the NP PP constructions and the semantic differences between the two constructions, the discussion in Lefebvre considers the fact that Haitian Creole has a lot more double object verbs (NP NP) than Fongbe. This problem is addressed from the point of view of the case-assigning properties of the verbs involved in the construction. 3.3.2.13 Summary. The data reported on in this chapter show that the bulk of the syntactic properties of Haitian verbs take on the properties of the substratum language rather than those of the superstratum language. Hence, the properties of Haitian BODY-state, WEATHER-, raising and control verbs contrast with those of French and pair with those of the substratum languages. Furthermore, both Haitian and Fongbe have a class of double-object verbs in contrast to French, which does not. This situation follows directly from the relexification hypo thesis. The selectional properties of Haitian reflexive verbs were shown to fol low from relexification followed by dialect levelling. The selectional properties of Haitian verbs that take overt/covert expletives do not perfectly match those of Fongbe; it is possible that dialect levelling has played a role in this case as well; but recall that the availability of a null expletive is a property of both Haitian and Fongbe, but not of French. The properties of the Haitian verb gen were hypothesised to result from an independent development within the creole. The subsets of verbs studied for case properties show that case-assigning properties constitute the syntactic properties that seem to be the most independent of the Creole's source languages. The case-assigning properties of double-object verbs were shown to provide further support for this claim. The lack of inherent object verbs in Haitian was attributed to the influence of French. The conclusion is thus that, aside from a few exceptions, the bulk of the syntactic properties of Haitian verbs reproduce those of its substratum languages. 3.3.3 Derivational affixes On the assumption that derivational affixes are listed in the lexicon as individual lexical entries that are minimally specified for categorial features and selectional and semantic properties (e.g. di Sciullo and Williams 1987; Lieber 1980, 1992; etc.) on the one hand, and on the hypothesis that relexification plays a central role in creole genesis, lexical entries of derivational affixes are expected to undergo relexification in a way similar to how nouns and verbs do. As is extensively argued for in Lefebvre (1998a: 303-334), the comparison of the derivational affixes of Haitian with those of its contributing languages does indeed show that their properties are quite straightforwardly derivable from the process of relexification. The paragraphs below briefly summarise this analysis.

THE RELEXEFICATION ACCOUNT OF CREOLE GENESIS

85

The inventory of the productive affixes of Haitian is as in Table 3.1. (Tests and methodologies for identifying a Creole's productive affixes are extensively discussed in Lefebvre 1998a: 303-312, and the references therein. See also section 10.1 of this book.) Table 3.1. The inventory of HC productive affixes Agentive suffix base output Attributive suffix base output Verbalising suffix base output Inversive/privative prefix base output Diminutive prefix base output Nominalising suffix base output Conversion base output Adverbial suffix base output Place of origin/residence suffixes base output Ordinal suffix base output

-è V N -è N N -e N V deV V tiN N -ay V N

V N/A -man A Adv -wal-yen N N -yèm Q A

(adapted from Lefebvre 1998a: 312) Why is the Haitian inventory of derivational affixes the size it is? Why does Haitian have the particular affixes that it has? The productive Haitian affixes with there semantically closest French equivalents are presented in Table 3.2. The French forms that are not in parentheses are those hypothesised to have provided the phonetic matrices from which the phonological representations of the Haitian affixes were assigned;

86

ISSUES IN THE STUDY OF PIDGIN AND CREOLE LANGUAGES

those in parentheses are forms that are synonymous, though they may differ as to subcategorisation and morphophonemic properties. Table 3.2. The HC affixes and their closest French corresponding forms Agentive suffix base output Attributive suffix base output Verbalising suffix base output Inversive/privative prefix base output Diminutive prefix base output Nominalising suffix base output Conversion base output Adverbial suffix base output Place of origin/residence suffixes base output Ordinal suffix base output

HAITIAN -è V N -è N N -e N V deV V tiN N -ay V N

V N/A -man A Adv -wal-yen N N -yèm Q A

FRENCH -CUT

V N -eur {-ard, -ier, -ien) N N -er, -é (-ir, -ifier, -iser) N/A V dé- (é-, in-, ir-) V/A V adjective petit [ti] (-et, -ot, -on) N/A N -age (-ion, -ment, -ance, -ure) V N - (-i, -é, -ert, -u) V N -ment A Adv -oisl-ien (-ais, -al, -and, -ain, -an) N N -ième Q A

(adapted from Lefebvre 1998a: 313-4) The data in Table 3.2 show that the derivational affixes of Haitian Creole all have at least one phonetically similar corresponding affix in French, except for the phonologically null affix (see below). But the most striking fact about the distribution in this table is that, in most cases, except for the agentive suffix and the phonologically null affix in cases of conversion, there are several French affixes corresponding to a single Haitian affix (see also Brousseau, Filipovich and Lefebvre 1989: 18). For example, while French has several overt affixes converting verbs into nouns, Haitian has only one; similarly, while French has

THE RELEXIFICATION ACCOUNT OF CREOLE GENESIS

87

several affixes designating a place of origin, Haitian has only two. Why did the other French affixes not make their way into Haitian? The Haitian affixes compare with the Fongbe ones (discussed in Lefebvre 1998a: 318-320, and the references therein), as in Table 3.3. Table 3.3. The HC affixes and their Fongbe corresponding forms HAITIAN Agentive suffix -è base V output N Attributive suffix -è base N output N Verbalising suffix -e base N output V Inversive prefix debase V output V Diminutive affix tibase N output N Nominalising suffix -ay base V output N Conversion base V output N/A Adverbial suffix -man base A Adv output Place of origin/residence suffixes -wal-yen base N output N Ordinal suffix -yèm base Q output A

FONGBE

-t V N

-n N N

màV/A V/A -Vi

N N copy prefix V N copy prefix V N/A

_ -tl-nui N N

-g Q A

(adapted from Lefebvre 1998a: 320-1) The Haitian and Fongbe derivational affixes in Table 3.3 differ in their phonological representation and they may also differ in their position with respect to the base they attach to. As we saw in section 3.1, these properties of the Haitian affixes appear to have been largely provided by the superstratum language. In spite of these differences, however, there is a striking resemblance between the two inventories: in most cases, there is a one-to-one correspondence

88

ISSUES IN THE STUDY OF PIDGIN AND CREOLE LANGUAGES

between the Haitian and Fongbe affixes. In both languages, there is one agen tive, one attributive, one inversive, one diminutive and one ordinal affix. It is a remarkable fact that, in both Haitian and Fongbe, there are exactly two suffixes referring to a place of origin/residence. Furthermore, Haitian -0 is involved in both nominal and adjectival conversion, two constructions which require the copy prefix in Fongbe. The correspondence between the Haitian and Fongbe affixes enumerated above contrasts with the Haitian/French data compared in Table 3.2, where it was shown that the majority of the Haitian affixes correspond to more than one affix in French. How can the correspondences between the Haitian and Fongbe lexical entries be accounted for? The similarity between the Haitian and Fongbe inventories in Table 3.3 can be accounted for in terms of the relexification hypothesis. The Haitian verbalising suffix -e, as well as the adverbial suffix -man have no counterpart in the substratum languages. They are analysed as innovations triggered by French morphology. (For a discussion of how relexification proceeds in the case of derivational affixes, of how the position of the morphological head is established in creole genesis, and of cases of dialect levelling in this area of the lexicon, see Lefebvre 1998a: 303-334, and the references therein. See also chapter 10 of this book.) Finally, in Lefebvre (1998a: 303-334; 403-407), it is extensively argued that the derivational affixes of Haitian are concatenated with their bases in a way which patterns on the substratum languages rather than the superstratum language. 3.3.4 Functional category lexical entries involved in nominal structure This section summarises the comparison of Haitian, French and Fongbe facts pertaining to the functional category lexical entries involved in nominal structure. The data in (67) provide an overview of French nominal structure. They show that, in this language, the definite, the possessive, and the demonstrative determiners all precede the head noun, and that there can be only one of these per noun phrase. Singular and plural forms are contrasted in (b) showing that plural is encoded in a bound morpheme in French. (67)

a.

He

mon ce

DEF

POSS DEM crab

crabe

FRENCH

THE RELEXIFICATION ACCOUNT OF CREOLE GENESIS

89

The Haitian and Fongbe nominal structures are illustrated in (68). In both languages, the determiners all follow the head noun. In both languages, a possessor phrase, a demonstrative term, the definite determiner and the plural marker may all co-occur within the same nominal structure. In both languages, the plural marker is an independent morpheme, as is shown in (68). (68) krab [mwen 0] sa a yo HAITIAN àsn [nyè ton] él  lé FONGBE crab me GEN DEM DEF PL 'these/those crabs of mine' (in question/that we know of) (=(l)in Lefebvre l998a:78) The Haitian and Fongbe nominal structures thus contrast in the same way with the French nominal structure with respect to word order, co-occurrence restrictions of determiners, and with respect to whether the plural marker is a free (in Haitian and Fongbe) or a bound (in French) morpheme. In the following sections, the properties of the definite determiner, the plural marker, the indefinite determiner, the deictic terms, and the case markers occurring within the noun phrase will be discussed in turn. 3.3.4.1 The definite determiner. The definite determiners of Haitian and Fongbe are shown in (69) and (70), respectively. (69) larivyè a river DEF 'the river' (in question/that we know of) (70)

vi  child DEF 'the child' (in question/that we know of)

HAITIAN

FONGBE

With the exception of their phonological representations, the properties of the definite determiners are the same in Haitian and in Fongbe; these properties contrast in a systematic way with those of the French definite determiner. These contrastive properties are summarised in (71) based on the detailed description in Lefebvre (1998a: 79-84).

90 (71)

ISSUES IN THE STUDY OF PIDGIN AND CREOLE LANGUAGES FRENCH

HAITIAN/FONGBE

[+ definite] determiner - pre-nominal - marked for gender and number - allomorphs: lellalles/l' - anaphoric and cataphoric - partitive du/des - obligatory with generic or mass nouns - no bare NPs - *Det [relative clause] N

[+ definite] determiner - post-nominal - unmarked for gender and number - allomorphs: la, a, an, nan, lanl,n - anaphoric - no partitive forms - impossible with generic or mass nouns - bare NPs - N [relative clause] Det

Moreover, the definite determiners involved in the Haitian and Fongbe nominal structures also play a crucial role in the clause structure of these two languages. In this case, the determiner has scope over the event denoted by the clause. This is exemplified in (72). (72)

a. b.

Li ríve a É wá  'He has arrived' (as expected/as we knew he would)

HAITIAN FONGBE

The French definite determiner plays no role at all in the clause structure. The determiner in the clause will be further discussed in section 3.3.5.9. Furthermore, as will be seen in section 3.4.6, the same determiner may occur in verb-doubling constructions in both Haitian and in Fongbe. These constructions have no equivalent in French. In Lefebvre (1998a: 82-84), it is argued that the postposed definite determiner of the substratum languages has been relabelled on the basis of the French deictic adverbial là occurring after noun phrases and clauses. It is also argued that if the two lexical entries that are associated in relexification share some properties, they are far from being identical. 3.3.4.2 The plural marker. A Haitian or Fongbe nominal structure may contain a noun followed by the plural marker only, as is shown in (73). In such a case, the structure is interpreted as definite. (73) krab yo àsn crab PL 'the crabs' *'(some) crabs '

le

HAITIAN FONGBE (=(31 ) in Lefebvre 1994a)

THE RELEXIFICATION ACCOUNT OF CREOLE GENESIS

91

Comparable data are impossible in French. The data in (74) show that Haitian and Fongbe both allow for bare NPs. (74) M' N'

achte krab.

HAITIAN

x

FONGBE

àsn.

I buy crab 'I bought (some) crabs.'

(=(32) in Lefebvre 1994a)

Bare NPs are not allowed in French. The data in (75) show that, in both Haitian and Fongbe, when the definite determiner and the plural marker co-occur within the same nominal structure, the definite determiner must precede the plural marker. (75) krab la yo / *yo a8 àsn  le / crab DEF PL PL DEF 'the crabs (in question)'



*le

HAITIAN FONGBE

(=(33) in Lefebvre 1994a)

In both languages, there is variation among speakers with respect to the possibility of co-occurrence of the determiner and the plural marker. Crucially, the patterns of variation are the same in both languages. Two slightly different grammars have been reported on in the literature. They are summarised in (76). (76)

HAITIAN

G1 where la and yo can co-occur (d'Ans 1968: 105; Faine 1937: 83; Fournier 1977: 43; Goodman 1964: 45; Joseph 1988: 201; Lefebvre and Massam 1988: 215; Ritter 1992: 207-209; Sylvain 1936: 55; Valdman 1978: 1994-1995) G 2 where la and yo cannot co-occur (DeGraff 1992b: 107; Joseph 1988: 201; Lumsden 1989: 65)

FONGBE where 6 and le can co-occur (Brousseau and Lumsden 1992: 22; Lefebvre 1998a: 85)

where  and le cannot co-occur (Agbidinoukoun 1991: 149)

In spite of their remarkable similarity, Haitian yo and Fongbe le differ in that yo, but not le, is also used as a third-person plural personal pronoun, as is shown in (77).

8 Note that in Haitian, the surface sequence krab yo a [crab 3pl DEF] is licit with the interpretation 'their crab' where yo is interpreted as the possessor (see section 3.3.1.2.1 on pronouns), rather than as the plural marker.

92 (77)

ISSUES IN THE STUDY OF PIDGIN AND CREOLE LANGUAGES a.

krab yo crab PL 'the crabs'

b.

yo pati 3rd.PL leave 'they left'

HAITIAN (=(28) in Lefebvre 1998a: 85)

In Fongbe, the third-person plural personal pronoun is expressed by a different morpheme, as shown in (78). (78)

a.

àson le crab PL 'the crabs'

b.

ye 3rd.PL leave 'they left'

yi

FONGBE

(=(29) in Lefebvre 1998a: 85)

In Lefebvre (1998a: 86-87), it is extensively argued that in spite of this difference, Haitian yo has been created through relexification and dialect level ling. In this analysis, the third person plural personal pronoun of the substratum languages was relexified on the basis of the strong pronominal French form eux, yielding yo (a predictable phonological derivation, as per the analysis in Brousseau in preparation). The use of this form was extended to the nominal structures following comparable lexical entries in other substratum languages of Haitian such as Ewe, where the morpheme wo, the third person plural pronoun, also encodes plural within the nominal structure. (For a theoretical account of the use of the same morpheme in these two environments, see Ritter 1992.) Speakers of languages of the Fongbe type who had two different morphemes, as in (78), had to conform to the dialect that had only one morpheme as a result of dialect levelling (for further discussion, see section 9.3). 3.3.4.3 The indefinite determiner. Fongbe has a postnominal indefinite determiner dé, as in àsn dé 'a crab'. Haitian has a prenommai indefinite determiner yon, as in yon krab 'a crab'. In Lefebvre (1998a: 88-89, and the references therein), it is shown that there was no French form available to relexify the substratum lexical entry. The Haitian prenommai numeral younn meaning 'one' is hypothesised to have been resyllabified as yon and reanalysed as a Haitian pernominal indefinite determiner. 3.3.4.4 The deictic terms. French has eleven deictic terms that can be involved in the nominal structure: ce(t) 'this/that', cette 'this/that', ces 'these/those', ça 'this/that', cela 'this/that', ceci 'this', celui 'this/that', ceux 'these/those', celle(s) 'this/that'; 'these/those', là 'there/here', ci 'here'. Speakers of Haitian divide into two groups with respect to the number of deictic terms: some speakers have two deictic terms and some have only one. These two lexicons will be discussed in turn. Haitian speakers who have two deictic terms in their lexicon pair with Fongbe speakers who also have two deictic terms. These are shown in (79).

THE RELEXIFICATION ACCOUNT OF CREOLE GENESIS (79)

HAITIAN

FONGBE

sa sila

(é)l (é)né

93

In Lefebvre (1997, 1998a: 89-101), it is extensively argued that the two Haitian terms do have the same distributional and syntactic properties as the Fongbe corresponding ones. For example, as is shown in (68) the postnominal demonstrative terms of Haitian and Fongbe may occur within the same nominal structure as a possessor, the definite determiner and the plural marker. In both languages, they may appear in contexts where we expect an NP. In (80) (adapted from (4) and (26) in Lefebvre 1997), they appear as part of an argument of the verb. (In examples (80) and (81), the deictic terms are glossed as DEIC and they are translated as 'this/that'. More precise semantic interpretation patterns are identified in (82)). (80)

M' wè sa / sila. N' m - él / éné I see one DEIC / DEIC 'I saw this/that one.'

HAITIAN FONGBE (=(4) in Lefebvre 1998b)

In (81) (adapted from (6) and (28) in Lefebvre 1997) they appear as part of the head of a relative clause. The examples also show that, when the head of the relative clause is plural, the deictic term is followed by the plural marker. (81)

a.

b.

-

sa yo 0-ki vini an. él le dé-é wá one DEIC PL OP-RES come DEF 'These/those ones who came.' sila yo 0-ki vini an. éné le dé-é wá one DEIC PL OP-RES come DEF 'These/those who came.'



HAITIAN FONGBE



HAITIAN FONGBE

In (80) and (81) the head of the nominal structure determined by the demonstrative terms is phonologically null. This null head corresponds to 'one' in English. Arguments supporting this analysis are provided in Lefebvre 1998a: 91-97). In Lefebvre (1998a: 91-97), it is extensively argued that the properties of the Haitian demonstrative terms discussed above differ from those of all the French deictic terms. It is proposed that the Haitian deictic terms have been created through relexification of the substratum ones on the basis of French ça and cela/celui-là yielding Haitian sa and sila, respectively (Lefebvre 1998a: 97-101). This proposal also accounts in a straightforward way for the three

94

ISSUES IN THE STUDY OF PIDGIN AND CREOLE LANGUAGES

semantic interpretation patterns of the two deictic terms identified in Lefebvre (2001a). These semantic patterns are reproduced in (82) as Gl, G2 and G3, where a is a variable that ranges over + or -. Crucially, these patterns of interpretation are identical for Haitian and Fongbe. (82)

a.

Gl sa [+proximate] sila [-proximate] HAITIAN G2 sa [α proximate] sila [-proximate] G3 sa [α proximate] sila [α proximate] Sources: Gl: Goodman (1964: 51), Tinelli (1970: 28). G2: Lefebvre (1997) [see also data in Etienne (1974) and in Sylvain (1936)]. G3: Férère (1974: 103), Joseph (1988), Valdman (1978: 194), Valdman et al. (1981) and my own fieldnotes. b.

Gl (é)lo [+proximate] (é)né [-proximate] FONGBE G2 (é)l [α proximate] (é)né [-proximate] G3 (é)l [α proximate] (é)né [α proximate] Sources: Gl: Anonymous (1983), Segurola (1963) and my own fieldnotes. G2: Lefebvre (1997). G3: My own fieldnotes. There are also Haitian speakers who have sa but not sila (e.g. Valdman 1996; Vilsaint 1992). As is discussed in detail in section 9.5, there are several substratum languages of Haitian that have only one general deictic term. There thus appears to be two Haitian lexicons with respect to demonstrative terms: one that has two terms which can be assigned three different patterns of interpretation, and one that has one term used as a general deictic term. Each of these lexicons corresponds to a substratum lexicon: one that has two terms and three patterns of interpretation, and one that has one term used as a general deictic term. 3.3.4.5 Case markers within the noun phrase. The Fongbe case marker tn in (68) is glossed as genitive. Arguments supporting the analysis of tn as a genitive rather than as an objective case marker are provided by Brousseau and Lumsden (1992). Since case markers have no semantic content, they cannot be relabelled. The Haitian phonologically null case corresponding to Fongbe tn in (68) is glossed as genitive on the basis of arguments provided in Lumsden (1991). Both languages also have an objective case, overt in Fongbe, covert in Haitian. These facts are in harmony with the general perspective adopted in our account of creole genesis are discussed at length in Lefebvre (1998a: 101-110, and the references therein). 3.3.4.6 Summary. The data discussed in this section show that the creators of Haitian did not perceive the functional categories involved in French nominal structure as such. They relabelled the determiner and the demonstrative terms of

THE RELEXrFICATION ACCOUNT OF CREOLE GENESIS

95

their own lexicon with phonetic strings corresponding to major lexical category items in French. The plural marker has come into the language through the relabelling of the third person plural pronoun. Case markers were assigned a phonologically null form. The history of the so-called indefinite marker yon does not follow this general pattern, however, as it appears to have developed though reanalysis from within the creole. 3.3.5. Functional category lexical entries involved in clause structure This section considers the functional categories involved in the structure of the clause. The tense, mood and aspect markers, the complementisers, the relative operator, the clausal conjunction, the focus marker, the marker of negation, the markers expressing the speakers point of view and the determiner in the clause will be discussed in turn. 3.3.5.1 The tense, mood and aspect markers. In Haitian Creole, the verb of a finite clause is invariant. In French, however, the verb of a finite clause obligatorily bears inflectional morphology encoding tense, mood, aspect, and person and number. None of the verbal morphology found in French has made its way into Haitian. Haitian follows the pattern of its West African (non-Bantu) substratum languages in having invariant bare verbs. In both Haitian and Fongbe, temporal relationships, mood and aspect are encoded by means of markers occurring between the subject and the verb. The inventory of the TMA markers9 of Haitian and Fongbe is provided in Table 3.4 (Bentolila 1971;Lefebvre 1996b, 1998a: 11-140).

9

The inventory of TMA markers in Haitian and Fongbe is established in Lefebvre (1996b) on the basis of syntactic tests which set the preverbal markers apart from modal and aspectual verbs. First, they all occur between the subject and the verb. Second, preverbal markers occurring in the same column in Table 3.4 are mutually exclusive, showing that they are in a paradigmatic relationship. Third, while modal verbs do allow for deletion of their VP complement, preverbal markers do not (for Haitian, see Koopman and Lefebvre 1982; Magloire-Holly 1982; Spears 1990; for Fongbe, see Lefebvre 1996b). Fourth, most of the preverbal markers in Table 3 4 have no meaning outside of the TMA system. Finally, the TMA markers may combine to form complex tenses.

96

ISSUES IN THE STUDY OF PIDGIN AND CREOLE LANGUAGES

Table 3.4. The inventory of TMA markers in Haitian and in Fongbe ANTERIOR

IRREALIS

Past/Past perfect H F te ko

Definite future H F ap ná

NON-COMPLETE Habitual H F — no

Imperfective H F ap dó...wè

Indefinite future H F a-va ná-wá Subjunctive H F pou ni

(=(115) in Lefebvre 1996b: 281) As can be seen from this table, the two inventories are remarkably similar. Both languages have a marker which encodes anteriority. Both lexically distinguish between definite and indefinite future. The definite future markers are used to convey the speaker's attitude that the event referred to by the clause will definitely take place in the near future. By contrast, the indefinite future markers are used to convey the speaker's opinion that the event referred to by the clause might eventually or potentially take place at an undetermined point in the future. The fact that speakers of Haitian distinguish between definite and indefinite future has been widely documented in the literature (see Spears 1990, and the references therein; Valdman 1970, 1978). For Fongbe, this distinction is pointed out in Anonymous (1983: V, 3). Both languages have a marker glossed as "subjunctive" for convenience. This term subsumes the three meanings of pou and ni respectively: both may be interpreted as 'must', 'should' or 'may'. Both languages have a form which encodes imperfective aspect. As can be seen in Table 3.4, there is a one-to-one correspondence between the preverbal markers in the two languages, except that Fongbe has one encoding the habitual aspect, and Haitian does not. As is shown in Lefebvre (1998a: 111-140) the phonological representations of the tense, mood and aspect markers of Haitian are derived from French periphrastic froms. For example, the anterior marker te is phonologically derived from the French auxiliary été, the imperfective marker ap is phonologically derived from the French periphrastic form après, the subjunctive marker pou from the French periphrastic form pour, and so on and so forth. In Lefebvre (1998a: 111-140), it is extensively argued, however, that the syntactic and semantic properties of the Haitian forms follow the details of the corresponding substratum lexical entries rather than those of the French

THE RELEXIFICATION ACCOUNT OF CREOLE GENESIS

97

forms from which they are phonologically derived. For example, the form of the Haitian indefinite future va is phonologically derived from the form of the French periphrastic future va. In French, however, the periphrastic future is used to encode definite future rather than indefinite future (Grevisse 1975: 731). The Haitian definite future marker is argued to have been created by the reanalysis of the clause initial adverb apre 'after' in much the same way as the Tok Pisin adverb baimbai 'after' (Sankoff and Laberge 1980). Three cases of relexification followed by levelling involving the anterior, the imperfective and the habitual markers are also discussed (Lefebvre 1998a: 127-129, 137-139). In both languages, complex tenses are formed by a combination of the preverbal markers rather than with auxiliary verbs as in French. For example, the combination of the markers of anteriority and of definite future yields a conditional interpretation of the clause as is shown in (83). Whether the conditional is assigned a present or a past interpretation is determined by the context. (83)

Mari te ap prepare pat. Mari ko ná dà wo. Mary ANT DEF.FUT prepare dough 'Mary would prepare dough.' 'Mary would have prepared dough.'

HAITIAN FONGBE (=(123) in Lefebvre 1996a)

Lefebvre (1996a) provides an exhaustive list of the complex tenses of Haitian and Fongbe showing that the range of complex tenses is the same for both languages. Both Haitian and Fongbe allow for bare sentences (that is, sentences in which there is no preverbal marker), in contrast to French which does not. (84)

Mari prepare pat la. Mari dà wo 6. Mary prepare dough DEF 'Mary has prepared the dough.'

HAITIAN FONGBE (=(56) and (110) in Lefebvre 1996a)

In both Haitian and Fongbe, the temporal interpretation of such sentences is computed from the various components that participate in establishing the aspectual properties of a clause (see Déchaine 1991 for Haitian). These components are the aspectual class of the verb (Lumsden 1995a), the definiteness of the direct object of the verb (Damoiseau 1988), and the definiteness of the subject (see Bentolila 1987). For example, while a clause containing a dynamic verb and a definite object is assigned a present perfect interpretation as in (84) above, a clause containing a dynamic verb and a nondefinite object is interpreted as past, as is illustrated in (85).

98 (85)

ISSUES IN THE STUDY OF PIDGIN AND CREOLE LANGUAGES Mari prepare pat. Mari dà w. Mary prepare dough 'Mary prepared dough.'

HAITIAN FONGBE (=(48) and (111) in Lefebvre 1996a)

It thus appears that the expression and interpretation of tense, mood and aspect in Haitian Creole follows the semantic and syntactic pattern of its substratum language rather than that of French, even though French has contributed the labels of the Haitian tense, mood and aspect markers. 3.3.5.2 Complementisers and complementiser-like forms. This section examines the forms introducing the tensed complements of verbs of the SAY- and WANTclasses. In Haitian, the complementiser introducing sentential complements of verbs of the SAY-class is phonologically null, as can be seen in (86). (86)

Lii kwè / panse li i / j refè. he believe/ think he cure 'He believes/thinks that he is cured.'

HAITIAN (=(25) in Sterlin 1988)

On the basis of binding and extraction facts, Sterlin (1988, 1989) argues that there must be a null complementiser introducing the embedded clause in (86). In French, the tensed complement of verbs such as croire 'to believe', dire 'to say', penser 'to think', etc., is introduced by the [+ tense] complementiser que 'that' (e.g. Kayne 1976; Milner 1978), requiring that the verb in the embedded clause be marked for indicative mood. (87)

Jean croit / dit I pense [CP que Marie est partie] FRENCH John believe/ say/ think COMP Mary left 'John believes/says/thinks that Mary left.' (=(55) in Lefebvre 1993a)

Haitian has no overt form corresponding to French que. This tells us that the creators of Haitian did not identify que as a [+ tense] complementiser. In Fongbe, the tensed complement of verbs of the SAY-class is introduced either by a null complementiser, as per the analysis in Kinyalolo (1993b), or by d (lit.: 'to say'), as per an analysis along the lines of Lord's (1976). These two analyses may be represented as (88a) and (88b), respectively. (88) a. b.

Koku di [(d [0 [Bàyi wá]]] FONGBE Koku believe say COMP Bayi come Koku di [d [Bàyi wá]] FONGBE Koku believe COMP Bayi come (=(6) in Lefebvre 1998a: 186)

THE RELEXIFICATION ACCOUNT OF CREOLE GENESIS

99

As is pointed out in Lefebvre (1998a: 186), these two analyses are not necessarily incompatible as they could be viewed as reflecting two competing dialects in the synchronic lexicon of Fongbe. From this perspective, the representation in (88a) would correspond to a conservative dialect, and that in (88b) to a more innovative dialect, where the serial verb (p 'to say' has been reanalysed as the phonological form of the previously null complementiser. The Haitian data compare with the Fongbe data as follows. First, unlike the innovative dialect of Fongbe and like the conservative one, Haitian has a phonologically null complementiser. However, unlike the conservative dialect, it does not have a serial verb meaning 'to say' in the construction under discussion. It thus appears that the difference between Haitian and the most conservative dialect of Fongbe resides not in the properties of the complementiser itself but rather in the availability of such a serial verb in the complement of SAY-class verbs. Possibly, the phonologically null complementiser of the conservative substratum lexicon was simply carried over into the Haitian Creole lexicon. (For a discussion of Saramaccan, which appears to reflect the more innovative Fongbe lexicon, see Lefebvre 1998a: 186-187.) Complements of verbs of the WANT-class and of a small class of adjectives such as 'good' in Haitian are introduced by pou, as is illustrated below (see Koopman and Lefebvre 1981, 1982; Lefebvre 1993a; Sterlin 1988, 1989). (89)

Yo te vle [pou m te antre nan troup Jakmèl] HAITIAN they ANT want COMP me ANT join in troops Jacmel 'They wanted me to join Jacmel's troops.' [Lit.: 'They wanted that I joined Jacmel's troops.'] (=(10) in Koopman and Lefebvre 1982)

(90)

Li bon [pou m t a pati] HAITIAN it good COMP I ANT IND.FUT leave 'It is good for me to leave.' [Lit.: 'It is good that I leave.'] (=(65) in Lefebvre 1993a)

Koopman and Lefebvre (1981, 1982) show that the complementiser pou is homophonous with the preposition pou which selects NP complements, as in (91), or purposive clauses as in (92). (91)

Pote sa pou mwen. bring this for me 'Bring this for me.'

HAITIAN (=(4) in Koopman and Lefebvre 1982)

100 (92)

ISSUES IN THE STUDY OF PIDGIN AND CREOLE LANGUAGES M te bezwen èskont sa a pou m te repati. HAITIAN I ANT need money this DEF for I ANT start.again 'I needed this money for a new start.' [Lit.: 'I needed this money so that I could start again.'] (=(6) in Koopman and Lefebvre 1982)

The complementiser pou is also homophonous with the mood marker of obligation pou discussed in section 3.3.5.1. (93)

Mari pou prepare pat. Mary SUB prepare dough 'Mary should prepare dough.'

HAITIAN

While the preposition pou derives its phonological form from the French preposition pour 'for' which selects NP complements, as well as purposive clausal complements, the Haitian mood marker pou derives its phonological form from the French form pour occurring in the periphrastic expression être pour 'to be about to'. However, and as is pointed out in Lefebvre (1993a: 118-119), in contrast to Haitian pou, French pour does not introduce complements of verbs of the WANT-class nor adjectives of the GOOD-class. In French, the tensed complements of verbs and adjectives of the WANT-class are introduced by the complementiser que bearing a special feature that Kayne (1976) represents as [+ F], standing for subjunctive mood. While the que selected by verbs of the SAY-class requires that the verb of the complement clause to be marked for indicative mood, the que selected by predicates of the WANT- and GOOD-classes requires the verb of the complement clause to be marked for subjunctive mood.10 But the Haitian complementiser pou does not derive its properties from the corresponding French complementiser que [+ F]. This raises the question of the source of the properties of Haitian complementiser pou. In Fongbe, verbs of the WANT-class and a small class of adjectives such as 'good' are introduced either by nú or by ni. The sentences in (94) and (95) show instantiations of the complementiser nú. (94) Un jló nú à ni wá. I want COMP 2nd SUB come T want you to come.' [Lit.: T want that you come.']

FONGBE (Anonymous 1983: X, 2)

10 For extensive discussions on the properties of the two French homophonous complementisers que, see Goldsmith 1978; Hirschbühler 1978; Kayne 1976, 1978; etc.

THE RELEXIFICATION ACCOUNT OF CREOLE GENESIS (95) É nú ùn ní yì. it be.good COMP 1st SUB leave 'It is good for me to leave.' [Lit.: 'It is good that I leave.']

101 FONGBE

(Anonymous 1983: X, 2)

The complementiser nú in (94) and (95) is homophonous with the preposition nú 'for' which selects either NP complements, or purposive clausal complements. The complement of the same predicates may also be introduced by ní. This is shown in (96) and (97). (96) Ùn jló ní à ní wá. I want COMP 2nd SUB come T want you to come.' [Lit.: T want that you come.']

FONGBE

(97) É ní ùn ní yì. it be.good COMP 1st SUB leave 'It is good for me to leave.' [Lit.: Tt is good that I leave.']

FONGBE (Anonymous 1983: X, 2)

The complementiser ní is homophonous with the mood marker ní discussed in section 3.3.5.1. According to the Fongbe speakers with whom I did fieldwork, the complementisers nú and ní are mutually interchangeable, that is, the selection of one or the other of these forms does not entail a semantic difference. Once again, the relexification hypothesis of creole genesis accounts in a straightforward way for the Haitian Creole data. In this view, the lexical entry copied from Fongbe nú, preposition and complementiser was relabelled as pou on the basis of the French preposition pour 'for' introducing nominal or clausal complements, yielding Haitian pou, preposition and complementiser. The lexical entry copied from Fongbe ní, mood marker and complementiser, was relabelled as pou on the basis of French pour occurring in the French periphrastic expression être pour 'to be about to', yielding Haitian pou preposition and complementiser. This is schematised in (98). (98)

Lexical entry preposition and complementiser mood marker and complementiser

HAITIAN pou pou

FONGBE nú ni

The two Haitian lexical entries signalled by pou are accidentally homophonous due to the superstratum forms that they were relabelled from: pour in both cases. The relexification hypothesis accounts in a straightforward way for the striking similarity between the properties of the substratum lexical entries and those of the corresponding ones in the creole. In this view, there is no need for recourse

102

ISSUES IN THE STUDY OF PIDGIN AND CREOLE LANGUAGES

to reanalysis (of the preposition as complementiser or of the mood marker as complementiser) as was previously proposed in Koopman and Lefebvre (1981). (For much further discussion see Lefebvre 1998a: 186-193.) 3.3.5.3 Complementisers or resumptives in the context of extracted subjects? Languages present subject/object asymmetries. They offer different strategies to rescue a sentence whose subject has been extracted. French has qui, a special form of the complementiser que (e.g. Moreau 1971; Kayne 1976; etc.). Fongbe has a resumptive pronoun in the extraction site: é in the singular and yé in the plural, as per the analysis in Law (1994a, 1994b). Haitian has the form ki phonologically derived from the French special form of the complementiser qui [ki]. (For numerous examples, see Lefebvre 1998a: 193-203.) The question is whether Haitian ki has the function of a complementiser (Koopman 1982a, 1982b) or that of a resumptive pronoun occurring in subject position (Law 1992, 1994b; Lumsden 1990; Manfredi 1993). In Lefebvre (1998a: 193-203), all the arguments supporting the above mentioned analyses are extensively layed out and evaluated. The conclusion of this study is that Haitian ki has the properties of a resumptive pronoun occurring in the position of extracted subjects thus following the syntactic pattern of the substratum languages. The fact that ki does not have the properties of French qui provides another piece of evidence showing that the creators of Haitian did not acquire these properties because they did not have enough exposure to French. 3.3.5.4 The nominal operator in relative and factive clauses. Like other Gbe languages, Fongbe has a lexical operator ɖêè which shows up in relative and factive clauses, as shown in (99) and (100), respectively. (99)

FONGBE house OP-RES in I sleep sleep LOC DEF 'The house in which I slept.' (=(14d) in Kinyalolo 1993a)

(100)

FONGBE arrive OP-RES John arrive DEF make.happy for mother GEN 'The fact that John arrived made his mother happy.' (=(3) in Lefebvre 1994b)

Kinyalolo (1993a) and Collins (1994) argue that this morpheme is an operator rather than a complementiser. The basis for their claim is that ɖêè can pied-pipe postpositions. Since complementisers cannot pied-pipe material and ɖêè does, ɖêè cannot be a complementiser. The alternative is that it is an operator occurring in specifier of CP. Collins further argues that it is a nominal operator, since it can only be coindexed with nominal phrases (thus excluding postpositional phrases).

THE RELEXIFICATION ACCOUNT OF CREOLE GENESIS

103

Operators have no semantic content. Since relabelling is semantically driven, we would expect operators to be assigned a null form at relabelling. This prediction is borne out as Haitian has no overt operator, as shown in (101) and (102). (101) Fiyi øi m sòti ak li i a. girl OP I go.out with her DEF The girl I went out with.' (102) Wá ɖé-é Jan wá ... Rive ø Jan rive a ... arrive OP-RES John arrive DEF 'The fact that John arrived ... '

HAITIAN (=(44) in Koopman 1982a) FONGBE HAITIAN (=(3) in Lefebvre 1994b)

Since the operator is null in this language, pied-piping phenomena of the type observed in Fongbe are not attested in Haitian. The crucial question, however, is whether this null operator is nominal. Koopman (1982a) provides extensive evidence that the Haitian operator can only be coindexed with noun phrases and not with prepositional phrases. In relative clauses involving a PP, the operator in the specifier of CP is coindexed with a resumptive pronoun in the complement position of a preposition within the relative clause, as shown in (101). These facts make sense only if the null operator is nominal. How does Haitian compare with French? In French, there is no overt operator. However, there appears to be a distinction between the properties of the operator in standard and popular French. Whereas in standard French the relative operator can be coindexed with a PP, in popular French, it cannot. This suggests that, whereas the relative operator in standard French is not nominal, it is nominal in popular French. As is extensively discussed in Bouchard (1982), popular French relative clauses were common in the variety of French spoken in the 17th century. Assuming an analysis where relative clauses and factive clauses require a phonologically null nominal operator in order to be interpreted, all three language varieties would be similar in having a nominal operator. While this operator is overt in Fongbe, it is covert in both popular French and Haitian. The following historical scenario is proposed in Lefebvre (1998a: 203-205). It is unlikely that the creators of Haitian acquired the properties of the French operator. It is assumed that they used the properties of their native lexical entry in creating the Haitian lexicon. Since operators have no semantic content, the original lexical entry could not be relabelled. By hypothesis, it was assigned a null form at relabelling. In this view, the Haitian lexical entry inherited the nominal property of the substratum lexical entry. Since the new nominal operator is phonologically null, it cannot pied-pipe lexical material, hence the discrepancy between Fongbe (99) and Haitian (101). By hypothesis, the first generation of

104

ISSUES IN THE STUDY OF PIDGIN AND CREOLE LANGUAGES

Haitian native speakers have deduced the nominal character of the null operator on the basis of the fact that they were not exposed to an operator relating a PP to an empty position in the relative clause. 3.3.5.5 Clausal conjunction. Haitian has a conjunction (e)pi 'and then' used to conjoin clauses, as shown in (103). (103) Jan pati' (e)pi Mari rive. John leave and.then Mary arrive 'John left and then Mary arrived.'

HAITIAN (=(70) in Lefebvre 1993a)

This conjunction derives its phonological representation from the French sequence of words et puis (lit.: 'and then'), pronounced [(e)pi], which is used in complementary distribution with et 'and' to conjoin clauses and noun phrases, as shown in (104). (104) a. b.

Jean John 'John Jean 'John

est parti et/(e)pi Marie est arrivée. FRENCH AUX leave and Mary AUX arrive left and Mary arrived.' (=(73) in Lefebvre 1993a) etl(e)pi Marie FRENCH and Mary' (=(74) in Lefebvre 1993a)

In contrast to French, the Haitian conjunction (e)pi cannot be used to conjoin noun phrases (see (105a)). Conjunction of noun phrases is achieved by adjoining a prepositional phrase to the first noun. This phrase is headed by the preposition ak or (kòl)ak (
*Jan (e)pi Mari [Lit.: 'John and.then Mary'] Jan (kòl-)ak Mari 'John and Mary'

HAITIAN HAITIAN

Once again, the above distribution finds a straightforward explanation when we examine comparable data from the substratum languages. Koopman (1986) notes that in West African languages different lexical items are used to coordinate clauses and NPs. In Fongbe, for example, the conjunction is used to coordinate clauses but not NPs, as shown in (106). (106) a. b.

Jan yì Mari wá. John leave and.then Mary arrive 'John left and then Mary arrived.' Van Mari [Lit.: 'John and then Mary']

FONGBE (=(70) in Lefebvre 1993a) FONGBE (=(71) in Lefebvre 1993a)

THE RELEXIFICATION ACCOUNT OF CREOLE GENESIS

105

As is the case in Haitian, NPs are coordinated by adjoining a prepositional phrase headed by (lit.: 'with.at') to the first noun. (107) Jan Mari (kpó) John with Mary with 'John and Mary'

FONGBE (=(72) in Lefebvre 1993a)

The properties of Haitian (e)pi 'and.then' and (kòl-)ak 'close.with' are derivable straightforwardly according to the relexification hypothesis. 'and.then' was relexified as (e)pi 'and.then' on the basis of French et puis 'and then' and was relexified as (kdl-)ak on the basis of French coller 'to be close to' and avec 'with'. (For an extensive discussion of conjunction in Fongbe with comparative data from Haitian, see Lefebvre in press). 3.3.5.6 The cleft marker. In Haitian clefts are introduced by se, phonologically derived from French c'est [se/sɛ]'it is'. Both forms are illustrated in (108). (108) a. b.

Se Jan Mari wè. HAITIAN C'est Jean que Marie a aperçu. FRENCH 'It is John that Mary caught sight of.' (=(72) in Lefebvre 1998a: 206)

In spite of their apparent similarity, se and c'est have quite different properties (e.g. Lumsden 1990; Déprez and Vinet 1991; DeGraff 1992b, 1992c), and they are assigned different analyses. It thus appears that, although French c'est is the source of the phonological representation of Haitian se, it did not provide its other properties. The closest Fongbe form to Haitian se is wέ, which also occurs in clefts, as in (109). (109) Mari Jan Mary it.is John catch sight of 'It is Mary that John caught sight of.'

FONGBE (=(74) in Lefebvre 1998a: 207)

Se and wέ differ in their distributional properties: se occurs at the beginning of the clefted constituent, whereas wέ occurs at the end of it. They also differ in their other properties (see Lefebvre 1998a: 206-208). It thus appears that, in this case, the properties of se were not provided by the substratum language. Presumably, the creators of Haitian who had a lexical entry like Fongbe wέ did not find an appropriate form with a suitable distribution in the superstratum language to relabel it. Given the methodology adopted for the comparative study,

106

ISSUES IN THE STUDY OF PIDGIN AND CREOLE LANGUAGES

Lefebvre (1998a: 208) concludes that the properties of se constitute an independent development (see also DeGraff 1992b, for a similar claim). 3.3.5.7 Negation markers. The Haitian negation marker pa is homophonous with French pas 'not'. However, although French pas obviously supplied the form of the Haitian negation marker, it did not contribute its other properties. One contrast noted by DeGraff (1993a) is that, while Haitian pa generally precedes the tense, mood and aspect markers, in French pas always occurs after the finite verb. This contrast is illustrated in (110) and (111). (110) a.

b. (111) a.

b.

Jan pa t' avale nan mache. HAITIAN John NEG ANT IND.FUT go in market 'John would not have gone to the market.' (=(la) in DeGraff 1993a) *Jan t' av ale pa nan mache (=(ld) in DeGraff 1993 a) Jean (ne) serait pas allé au marché. FRENCH John (ne) would.be pas gone to.the market 'John would not have gone to the market.' (=(2a) in DeGraff 1993a) Jean n' ira pas au cinéma. FRENCH John (n') go.FUT pas to.the movies 'John will not go to the movies.' (=(2c) in DeGraff 1993a)

Another contrast noted by DeGraff (1993a) is that, whereas Haitian pa must occur between the subject and the verb, French pas may, in some contexts, occur at the periphery of the clause that it modifies. Compare (112) and (113). (112) Bouki fait le clown pour pas qu' ils s'ennuient. FRENCH Bouki makes the clown for pas that 3pl bore + REF 'Bouki is clowning around so that they don't get bored.' (=(1la) in DeGraff 1993a) (113) *Bouki ap fè komik pou pa yo anniye HAITIAN Bouki IMP make clown for pa they bore (=(11b) in DeGraff 1993a) A third difference is that French pas, but not Haitian pa, may occur in nominal structures, as shown in (114) and (115). (114) Voilà un type pas bête. there a fellow pas stupid 'There goes a man who is not stupid.'

FRENCH (=(12a) in DeGraff 1993a)

THE RELEXIFICATION ACCOUNT OF CREOLE GENESIS (115) *Men yon mounn pa sot here/there.is a fellow pa stupid 'There goes a man who is not stupid.'

107 HAITIAN

(=(12b) in DeGraff 1993a)

A fourth difference, discussed at length by both DeGraff (1993a) and Déprez (1999), has to do with the way Haitian pa and French pas interact with negative quantifiers. These data argue that, although French pas contributed the form of Haitian pa, it did not contribute its other properties. DeGraff (1993a) further claims that Haitian pa actually shares properties with French ne (see (111)). A comparison of Haitian pa with corresponding morphemes in the substratum languages reveals the source of the properties of pa. All Gbe languages have a negation marker that occurs between the subject and the verb (Hazoumê 1990). In Fongbe, this marker is mà. As is the case with Haitian pa, this marker generally precedes the tense, mood and aspect markers. Compare (116) with (110a). (116)

FONGBE Koku NEG SUB go market in 'Koku does not have to go to the market.' (=(85) in Lefebvre 1998a: 210)

In Haitian (for a subset of speakers) and in Fongbe, the mood markers can also precede palmà, as shown in (117). Note the effect of word order on the interpretation of the sentence. (117) a.

b.

Man pa pou prepare pat. HAITIAN Mari mà FONGBE Mary NEG SUB prepare dough 'Mary does not have to prepare dough.' (=(86a) in Lefebvre 1998a: 210) Man pou pa prepare pat. HAITIAN Mari ní mà FONGBE Mary SUB NEG prepare dough 'Mary should not prepare dough.' (=(132) in Lefebvre 1996a)

In Lefebvre (1998a: 210-211), it is further shown that like Haitian pa, Fongbe mà cannot occur at the periphery of the clause that it modifies, it is not allowed in nominal structures, and it interacts with negative quantifiers in a way similar to Haitian pa. On the basis of these facts, Lefebvre draws a twofold conclusion: Fongbe mà was relexified as pa on the basis of the French negation adverb pas; French ne was not identified as such by the creators of the creole and therefore, it did not enter Haitian Creole.

108

ISSUES IN THE STUDY OF PIDGIN AND CREOLE LANGUAGES

3.3.5.8 Markers expressing the speaker's point of view. Fongbe has a paradigm of functional items whose function is to express the speaker's point of view on the proposition. In the literature, lexical items of this type are referred to as evidentials or validators. Fongbe has three of these markers: the yes-no question marker à, the negative marker ä and the marker of insistence ó (see da Cruz 1994). All three occur only at the end of clauses. Evidentials are not part of the French lexicon. Joseph (1995) argues that Haitian Creole has a marker of insistence that has the properties of Fongbe ó. He proposes that ó has been relexified on the basis of French non. The bulk of the properties of French non, however, are not associated with Haitian non. This paradigm of lexical items is extensively discussed in Lefebvre (1998a: 213-217), where it is also shown that the substratum question and negative markers have not been relexified due to lack of available appropriate material in the superstratum language. 3.3.5.9 The determiner in the clause. The definite determiner found in the nominal structure of Haitian and Fongbe also plays a central role in the structure of the Haitian and Fongbe clause (see Lefebvre 1982, 1991b, 1992, 1996b, 1998b; Lefebvre and Massam 1988; Law and Lefebvre 1995). As shown in (118), when the determiner occurs in the context of a clause, it may be assigned three slightly different interpretations. (118) Moun nan kraze manchinn nan an. HAITIAN1 Súnù FONGBE1 man DEF destroy car DEF DEF a. 'The man destroyed the car (as was said earlier).' b. 'The man has destroyed the car, as we knew he would.' c. 'The man has destroyed the car, as we knew it would be destroyed.' (=(2) in Lefebvre 1998b) In (118a), the determiner asserts the content of the proposition, relating it to something that has been said earlier in the conversation. In (118b) and (118c), the determiner identifies an event that is already part of the shared knowledge of the participants. It literally means 'this event in question/this event that we know of'. The determiner with this meaning has been referred to in earlier work as the event determiner (e.g. Lefebvre 1992). As such, it may trigger an interpretation which is subject-oriented, as in (118b), or object-oriented, as in (118c). Lefebvre (1998b) proposes that the clausal determiner may head one of four functional category projections in the clause. The position of the determiner in the syntactic tree determines its scope and therefore its specific interpretations. In Lefebvre (1998a: 219-248) it is shown that the properties of the determiner in the clause are the same in both languages. As an assertive marker, the determiner in the clause interacts with the evidential markers discussed in

THE RELEXIFICATION ACCOUNT OF CREOLE GENESIS

109

section 3.3.5.9. As an event determiner, the determiner in the clause interacts with the aspectual properties of the clause determined by the aspectual class of the verb, the overt manifestation of aspect, etc. In this case, it is licensed by the definiteness of the two arguments that delimit the event denoted by the clause: the subject and the affected object. Furthermore, in both languages, the surface distribution of the determiner in the clause is constrained by the same factors. Finally, and as is documented in detail in Lefebvre (1998a: 119-148), there is variation among speakers in each language. In both languages three patterns of variation have been identifed and they appear to be the same for both languages. The examples in (118) illustrate one of the three patterns. This finds no parallel in French, for, in this language, the determiner plays no role at all in clause structure. 3.3.5.10 Summary. The bulk of the properties of the functional lexical items involved in clause structure follows rather straightforwardly from the theory of creole genesis outlined in section 3.1. The inventory and properties of the tense, mood and aspect markers of Haitian, though phonologically derived from French periphrastic expressions, correspond to those of Fongbe (with one case of reanalysis and a few cases of dialect levelling). The null complementiser introducing complements of verbs of the SAY-class in Fongbe has a null counterpart in Haitian. In both Haitian and Fongbe, there are two underspecified lexical entries: one which can be used as mood marker and complementiser, and another which serves as preposition and complementiser. Haitian and Fongbe are also similar in having a resumptive pronoun in the basic position of extracted subjects, unlike French, which has a special form of the complementiser that licences the empty subject position. In both Haitian and Fongbe, the conjunction used to conjoin clauses cannot be used to conjoin NPs, in contrast to French where the same lexical item can conjoin both clauses and NPs. The negation marker in Haitian was argued to have the same semantic and distributional properties as the negation markers of the substratum languages, but not the French adverbial form from which it was phonologically derived. The lexical operator of Fongbe could not be relabelled because it has no semantic content. This lexical entry was assigned a null form in the Haitian lexicon. The interrogative marker à and the negative marker ä were not relexified because there were no available forms in the superstratum language to provide them with a new phonological representation. The Haitian marker of insistence non was shown to have the same properties as Fongbe ó. These data show that, as in the substratum languages, and in contrast to French, Haitian has grammatical markers which express the speaker's point of view on the proposition. The determiner which plays a role in nominal structure also plays a role in the clause structure. Again, this unites Haitian and Fongbe against French, where the

110

ISSUES IN THE STUDY OF PIDGIN AND CREOLE LANGUAGES

determiner occurring in a nominal strcuture plays no role at all in the structure of the clause. The sole Haitian lexical entry playing a role in the clause structure that appears to have had an independent developement is se. Indeed, its properties correspond to those of neither the substratum nor the superstratum language. So, for this area of the lexicon as well, we have to conclude that the relexification account of creole genesis accounts for the bulk of the lexical items involved. Appendix 2 provides an overview of the types of Haitian Creole lexical entries with respect to origin. 3.4. Parameters By hypothesis, the creators of the creole use the parametric values of their own grammar in assigning a value to the parameters of the language that they are creating. This hypothesis predicts that, where the parametric values of the substratum and superstratum differ, the creole should have the same value as the substratum languages. In the case of Haitian Creole, with one exception, this prediction is borne out. As will be seen below, this is largely due to the fact that the properties of the functional categories have been reproduced in the creole through relexification. The following parameters will be discussed: availability of null subjects, verb raising, serial verbs and double-objects, negative quantifiers and verb-doubling phenomena.11 The content of this section summarises chapter 12 of Lefebvre (1998a: 349-375). 3.4.1 The null subject parameter One of the parametric options of UG relates to whether null subjects are available in particular languages. In point of fact, this parameter represents the remains of the former PRO-DROP parameter expressed in Chomsky (1981) (e.g. Bennis 1982; Hulk 1986; Law 1992; Safir 1982). In recent literature, it has been proposed that languages with syntactic clitics should be considered null subject languages (e.g. Jaeggli 1984; Hulk 1986; Roberge 1990). The theory goes as follows: subject clitics are not generated in NP positions but in a functional category projection (INFL(ection) or AGR(eement)) as the spelling-out of person, number, gender and case features. In languages which have syntactic clitics, the subject position is thus phonologically null, but it is bound by the clitic. In this view, both French and Fongbe would be null subject languages, since both languages have syntactic clitics, as is extensively argued for in Lefebvre (1998a: 148-157).

11 The discussion is based on a comparison of Haitian, French and Fongbe (and other Gbe and Kwa languages). The Bantu languages have parametric values which are sometimes quite different from those of the Kwa languages. They are not considered here (for a discussion on this issue, see Lefebvre 1998a: 390-393).

THE RELEXIFICATION ACCOUNT OF CREOLE GENESIS

111

In the recent literature on Haitian Creole, there has been some debate as to whether Haitian is a null subject language (see Cadely 1994; DeGraff 1992a, 1992b, 1992d, 1993b, 1996; Déprez 1992a; Law 1992). This debate depends on whether Haitian has null subjects of the type we find, for example, in Italian. DeGraff (1992a) claims that there are empty subjects in Haitian. Déprez (1992a) and Cadely (1994) argue against this position. They both argue that Haitian clitics are not syntactic but phonological. Having evaluated the arguments presented to support each of these analyses, I also conclude that Haitian is not a null subject language. Both sets of arguments are fully presented and extensively discussed in Lefebvre (1998a: 148-157). Assuming the analysis whereby the availability of syntactic clitics defines a language as a null subject language, we have to conclude that Haitian differs from both of its source languages on this parameter. While both French and Fongbe have a positive value for the null subject parameter, Haitian has a negative one. Thus, in this case, it appears that the creators of Haitian had to reset the value of the original parameter. This situation is a consequence of the fact that they have abandoned the syntactic clitics of their original lexicon. As is extensively discussed in Lefebvre (1998a: 148-157), based on Brousseau (1995a), it is shown that syntactic clitics were not reproduced in the early creole due to the way relexification proceeds in creole genesis. As a consequence of this situation, it is likely that the first generation of Haitian native speakers assigned the null subject parameter a negative value since they were exposed only to strong subject pronouns. In terms of the markedness issue, this is extremely interesting. On the basis of work by Hyams (1986, 1987), DeGraff (1992a) points out that availability of null subjects is the unmarked option of UG. If this is correct, while both the substratum and the superstratum languages of Haitian present the unmarked option of this parameter, Haitian exemplifies the marked one. This is a major drawback for theories advocating that creole languages systematically present the unmarked parametric options of Universal Grammar (e.g. Bickerton 1984). 3.4.2 Verb raising In recent literature, it has been proposed that languages vary based on whether they allow verb raising (Chomsky 1981; Pollock 1989, and related literature). On the basis of facts involving, among other phenomena, negative placement, question formation and adverb placement, Pollock (1989) argues that while French has verb raising, English does not. In Pollock's (1989) analysis, this cluster of differential properties between English and French can be accounted for by a parametric difference between the two languages, depending on whether or not the language allows verb raising. In French, the verb must raise to a higher position in the syntactic tree (from V through AGR to tense, or even to

112

ISSUESINTHE STUDY OF PIDGIN AND CREOLE LANGUAGES

CP). By contrast, verb raising is not available in English, and hence, the verb stays in its basic position within the VP. Pollock's analysis also captures the relationship between the availability of verb raising and rich verbal morphology in a given language: only languages with rich verbal morphology, such as French, have verb raising. It is argued that this inflectional morphology is picked up by the verb as the verb moves through AGR to tense. Both Haitian and Fongbe contrast with French with respect to this parametric option. Unlike French, neither Haitian nor Fongbe has inflectional morphology (see section 3.3.5.1). Furthermore, neither Haitian nor Fongbe present any of the characteristics of the verb raising languages (see Lefebvre 1998a: 351-355), a conclusion which accords with DeGraff (1992b) and Avolonto (1992), for Haitian and Fongbe, respectively. This contrasts with French, which is a verb raising language par excellence. Thus, for this parameter, Haitian has the same value as the Gbe languages, and the Kwa languages more generally (e.g. Givón 1971; Baker 1991).12 It thus appears that the value of the verb raising parameter in Haitian pairs with Fongbe and differs from French. As is pointed out in Lefebvre (1998a: 355), it appears that the first generation of Haitian native speakers were able to identify the properties of INFL and AGR in the language they were exposed to on the basis of the primary data that were submitted to them. On the basis of these properties, they deduced that verb raising is not available in that language. 3.4.3 Serial verbs Like several West African languages, Haitian Creole has serial verbs, as shown in (119). (119) É àxì Li pran crab ale nan mache. 3rd take crab go in market in 'He brought the crab to the market.'

mέ.

FONGBE HAITIAN (=(8) in Lefebvre 1986)

In contrast to Haitian and Fongbe, French does not have serial verbs. In recent literature, several parameters/correlations have been proposed to account for the availability of serial verbs in particular languages. Among the proposals that have been made, two are borne out by the data. (For a discussion of the other proposals, see Lefebvre 1998a: 355-357.) A first proposal is that there is a correlation between the availability of verb serialisation in particular 12 On the basis of facts involving adverb placement in Haitian and Fongbe, DeGraff (1994) challenges the conclusion that Haitian and Fongbe are alike with respect to verb raising. A rebuttal of his analysis can be found in Lefebvre (1998a: 353-355).

THE RELEXIFICATION ACCOUNT OF CREOLE GENESIS

113

grammars and the lack of derivational verbal morphology (e.g. Baker 1991: 79). This correlation holds true for verbs in the Caribbean creoles and in West African languages, which are largely mono-morphemic (see Muysken 1988d). Data from the three languages under comparison here also support this claim. As was shown in section 3.3.3, French has many derivational affixes which modify the meaning of base verbs. For example, the base verb porter 'to carry' is part of the derived verb ap-porter 'to bring'. The latter concept is expressed by a serial verb construction in both Haitian and Fongbe, as is shown in (119). Both Haitian and Fongbe lack derivational affixes of the type we find in French. Hence, this correlation is supported by data drawn from the three languages being examined here. This correlation would account for the fact that, while French has no serial verbs, Haitian and Fongbe do have this construction. The correlation has further been extended to the lack of inflectional verbal morphology. In this view, the availability of verb serialisation correlates with the lack of inflectional morphology, and hence, with the absence of verb raising to INFL, as discussed in section 3.4.3 (e.g. Baker and Stewart 1996; Déchaine 1993; Muysken 1988d). This correlation accounts for the differential properties of the three languages under comparison here. French has inflectional morphology (see section 3.3.5.1) that the verb picks up while raising to INFL (see section 3.4.2). By contrast, Haitian and Fongbe do not have inflectional morphology (see section 3.3.5.1) and verb raising is not available in these languages (see section 3.4.2). In French, there are no serial verbs whereas in Haitian and Fongbe this option is available. According to the correlation presented above, the creators of Haitian, who were native speakers of languages of the Kwa family, kept the parametric value of their original language in creating the creole: they did not have derivational verbal morphology, and they did not have verb raising to INFL due to lack of inflectional morphology, and hence verb serialisation was available to them. The first generation of native speakers of Haitian would have identified the absence of verb raising in the language that they were presented with. Having identified this parametric value, they deduced the availability of verb serialisation in the grammar (see Lefebvre 1998a: 355-357). 3.4.4 The double-object construction As we saw in section 3.3.2.12, in contrast to French, both Haitian and Fongbe have the recipient-theme construction (NPNP). Among the numerous correlations proposed to account for the availability of the double object construction (all discussed in Lefebvre 1998a: 357-360), there is only one that is supported by the three languages under scrutiny here. Johnson (1991) proposes a direct correlation between the availability of the double-object construction and the availability of structural genitive case (e.g. 's in English) in nominal structures.

114

ISSUES IN THE STUDY OF PIDGIN AND CREOLE LANGUAGES

According to this proposal, the double-object construction, as in (120a), is available in a given grammar because structural genitive case, exhibited in (120b), is also available in the nominal structure of that grammar. (120) a. John gave Mary a book. b. Mary's book

ENGLISH

The motivation for Johnson's proposal is the claim that the two NPs involved in the double-object construction are in a possession relationship which parallels the relationship observed in nominal structures between the possessed and the possessor marked for genitive case. This correlation is borne out by data from the three languages under comparison. The double-object construction is not available in French because genitive case is not available in French nominal structures. The double-object construction is available in Haitian and Fongbe because, as we saw in section 3.3.4.5, in both Fongbe and Haitian nominal structures, genitive case is available. The correlation proposed in Johnson (1991) between the availability of the double-object construction and the availability of structural genitive case in nominal structures is thus supported by these data. In Lefebvre (1998a: 359), it is hypothesised that the creators of the creole used their knowledge of their own grammars and lexicons in setting the value of the parametric option which allows for double-object constructions. They had a genitive construction which they reproduced in the creole. This allowed them to have the double-object construction, which they also reproduced in the creole. The first generation of Haitian native speakers identified the genitive case in the nominal structure of the language they were presented with. On the basis of this property, they deduced the availability of the double-object construction in that language. 3.4.5 The interpretation of negative quantifiers Haitian negative quantifiers derive their phonological form from French phonetic sequences; for example, pèsonn is phonologically derived from French personne 'nobody', and anyen from French rien 'nothing'. However, the properties of these quantifiers are not derived from the corresponding French forms. First, DeGraff (1993 a: 67) points out that negative quantifiers interact differently with Haitian pa than with French pas. Second, he observes that, in French, "cooccurring negative elements cancel each other, giving rise to a net positive statement." This contrasts with Haitian where the two negative elements "are immediately construed as net negative statements." Third, Déprez (1999) points out that, in Haitian, negative quantifiers usually require the presence of a negative marker. This is shown in (121) where pa must occur.

THE RELEXIFICATION ACCOUNT OF CREOLE GENESIS

(121) a. b.

115

M *(pa) te wè pèsonn / anyen. HAITIAN I not ANT see no one / nothing 'I did not see anyone/anything.' (=(la) in Déprez 1999) Pèsonn *(pa) rive. HAITIAN 'No one arrived.' (=(2) in Déprez 1999)

As noted by Déprez (1999), this contrasts with standard French, where pas cannot occur in this context. The Haitian data also contrast with popular French (Déprez 1999 and Lefebvre 1998a: 79-84). Déprez (1999) discusses several other differences between the two languages. She concludes that the properties of the Haitian negative quantifiers cannot be attributed to French. On the basis of very careful and thorough argumentation, Déprez (1999) proposes accounting for the difference between Haitian and French in terms of their determiner systems. French does not have bare NPs (see section 3.3.4) but it has a partitive determiner de, du and des. Based upon Déprez's account, French negative quantifiers behave like indefinite determiners (or numerals) with empty nouns. By contrast, Haitian has bare NPs (see section 3.3.4), and negative quantifiers are nouns with empty Ds. These two structures are illustrated in (122) (adapted from (93) in Déprez 1999).

In addition, Déprez (1999) presents extensive arguments showing that 17th century French is not the source of bare NPs in Haitian Creole, a conclusion which is in agreement with my own (see Lefebvre 1998a: 79-89). In Fongbe, the negative quantifiers are mέtí 'nobody' and nŭtí 'nothing' corresponding to Haitian pèsonn and anyen, respectively. In Lefebvre (1998a: 360-363) it is shown that the negative quantifiers co-occur with negative or negation markers in sentences that are interpreted as negative statements. This contrasts with French but parallels the Haitian data. Furthermore, as is the case in Haitian, a clause containing a negative quantifier requires the presence of a negation or negative marker. According to Déprez's general proposal based on Haitian, Fongbe negative quantifiers would thus be like Haitian negative quantifiers, that is, NPs rather than determiners. Interestingly enough, and as we saw in section 3.3.4, Fongbe, like Haitian, allows for bare NPs. The Fongbe data thus appear to provide independent support for the formulation of the parameter proposed in Déprez (1999). On the basis of this comparison, Lefebvre

116

ISSUES IN THE STUDY OF PIDGIN AND CREOLE LANGUAGES

concludes that, although the phonological representations of the negative quantifiers in Haitian are derived from French, their semantic and syntactic properties are derived from those of the corresponding lexical entries in substratum languages such as Fongbe. Using the parameter defined by Déprez (1999), Lefebvre (1998a: 362) proposes to account for the history of the Haitian facts as follows. The forms pèsonn and anyen were incorporated into the early Haitian lexicon as nouns rather than as determiners. The first generation of Haitian native speakers encountered bare NPs in the language they were exposed to and deduced that negative quantifiers were NPs, rather than determiners, in this language. 3.4.6 Verb-doubling phenomena Verb-doubling phenomena are involved in four constructions which contain what looks like an exact copy of the predicate (henceforth "the copy"). Koopman (1986) points out that constructions involving a copy of the verb are attested in Haitian and in West African languages but not in French. Clauses containing what looks like a copy of the predicate involve four constructions: temporal adverbial, as in (123), causal adverbial, as in (124), factive clauses, as in (125) and the predicate cleft construction, as in (126). In the examples below, the first occurrence of the verb is an exact replica of the second one. (123) Temporal adverbial Wá Jan wá (tróló) Mari yì. FONGBE Rive Jan rive (epi) Mari pati HAITIAN arrive John arrive as.soon.as and Mary leave 'As soon as John arrived, Mary left.' (=(1) in Lefebvre 1994b) (124) Causal adverbial Wá Jan wá wútú Mari yì. Rive Jan rive Mari pati. arrive John arrive cause Mary leave 'Because John arrived, Mary left.'

FONGBE HAITIAN (=(2) in Lefebvre 1994b)

(125) Factive Wá ɖè-é Jan wá víví nú Rive ø Jan rive a, fè arrive OP-RES John arrive DEF make(-happy) for

FONGBE HAITIAN

manman li kòntan. mother GEN happy 'The fact that John arrived made his mother happy.' (=(3) in Lefebvre 1994b)

THE RELEXIFICATION ACCOUNT OF CREOLE GENESIS

(126) Predicate cleft Wá wέ Jan wá. Se rive Jan rive. it.is arrive it.is John arrive 'It is arrive that John did.' (not e.g. leave)

117

FONGBE HAITIAN (=(4) in Lefebvre 1994b)

The informants whose data are reported on in Lefebvre (1990), and Law and Lefebvre (1995) allow various contrastive interpretations of the clefted constituents. In the above examples, the contrastive reading relates to the V alone. In the examples below, even though the clefted constituent consists only of the copy of the verb, the contrastive reading involves the whole VP. (127) a.

b.

Se monje Jan monje pen an. HAITIAN it.is eat John eat bread DEF 'It is eat the bread that John did.' (not e.g. drink the water) (=(44) in Lefebvre 1990) FONGBE hit it.is Asiba hit Koku 'It is hit Koku that Asiba did.' (not e.g. kill Sika) (=(66) in Law and Lefebvre 1995)

Likewise, for these speakers, when the delimiting object is clefted, the contrastive reading may bear on this argument alone or on the whole VP, as shown in (128). (128) a.

Se pen an Jan monje. HAITIAN it.is bread DEF John eat 'It is the bread that John ate.' (not e.g. the meat) or 'It is eat the bread that John did.' (not e.g. drink the water) (=(53) in Lefebvre 1990) b. FONGBE car DEF it.is man a destroy 'It is the car that a man destroyed.' (not e.g. the bicycle) or 'It is destroy the car that a man did.' (not e.g. build the house) (=(72) in Law and Lefebvre 1995)

The semantic interpretation facts in (127) and (128) are remarkable and they show a striking parallel between the grammars of these subsets of Haitian and Fongbe speakers. Various accounts of these facts may be found in Lefebvre (1990), Larson and Lefebvre (1991), Collins (1994) and Law and Lefebvre (1995). Although these constructions differ in their semantic and syntactic properties (see Lefebvre 1998a: 363-374), they all share the fact that they contain a copy of the verb, and that, unlike deverbal nominais, the copy is

118

ISSUES IN THE STUDY OF PIDGIN AND CREOLE LANGUAGES

deprived of an argument structure. For a subset of both Haitian and Fongbe speakers (identified below as Haitian1 and Fongbe1), the copy can be followed by the determiner which otherwise occurs in nominal structures (see section 3.3.4.1), as is shown in (129)- (132). (129) Temporal adverbial Wá Jan wá (tróló) Mari yì. FONGBE1 Rive a Jan rive (epi) Mari pati. HAITIAN1 arrive DEF John arrive as.soon.as and Mary leave 'As soon as John arrived (as we knew he would), Mary left.' (=(19) in Lefebvre 1994b) ( 130) Causal adverbial Wá 6 Jan wá wútú Mari yì. FONGBE1 Rive a Jan rive Mari pati. HAITIAN1 arrive DEF John arrive cause Mary leave 'Because John arrived (as we knew he would), Mary left.' (=(20) in Lefebvre l994b) (131)Factive nú Rive a 0 Jan rive a, fè arrive DEF OP-RES John arrive DEF make(-happy) for

FONGBE1 HAITIAN1

manman li kontan. mother 3rd GEN happy 'The fact that John arrived (as expected) made his mother happy.' The (very) fact that John arrived made his mother happy.' (=(21) in Lefebvre l994b) (132) Predicate cleft Yì wέ Jan yì. FONGBE1 Se ale a Jan ale. HAITIAN1 it.is leave DEF it.is John leave 'It is leave (as expected) that John did.' (not e.g. stay home) (=(22) in Lefebvre 1994b) In contrast to speakers of Haitian1, speakers of what I will call Haitian2 accept the determiner only at the end of the clause containing the copy. The judgements of these speakers are illustrated in (133)—(136). (Similar judgements are also reported in Lefebvre and Ritter 1993.) (133) Temporal adverbial Rive Jan rive a (epi) Mari pati. arrive John arrive DEF and Mary leave 'As soon as John arrived (as we knew he would), Mary left.'

HAITIAN2

THE RELEXIFICATION ACCOUNT OF CREOLE GENESIS

(134) Causal adverbial Rive Jan rive a Mari pati arrive John arrive DEF Mary leave 'Because John arrived (as we knew he would), Mary left.'

119

HAITIAN2

(135) Factive Rive ø Jan rive a, fè manman li kontan. HAITIAN 2 arrive OP John arrive DEF make(-happy) mother his happy 'The fact that John arrived (as expected) made his mother happy.' 'The (very) fact that John arrived made his mother happy.' (136) Predicate cleft Se ale Jan ale a. it.is leave John leave DEF 'It is leave (as expected) that John did.' (not e.g. stay home)

HAITIAN2

Some speakers of the first group ((129)-(132)) even accept sentences where the event determiner occurs both with the copy (as in (129)-(132)) and at the end of the clause containing the copy (as in (133)—(136)). The variation observed among speakers is akin to the variation observed between West African languages with respect to whether they allow determiners to appear immediately after the head of a relative clause or a factive construction or at the end of the clause (for an extensive discussion of these facts, see Collins 1994). The additional data in (129)-(132) further illustrate the parallelism between the grammars of Haitian and Fongbe. Availability of verb-doubling phenomena in particular grammars has received some attention (see Lefebvre 1998a: 371-373). For the purpose of the present discussion, I will assume the proposal that availability of verb-doubling phenomena is linked to the properties of the determiner system. In this analysis, the properties that distinguish the Fongbe/Haitian definite determiner from the French determiner may account for the fact that, in the former type of language, the verb copy is available while, in the latter, it is not. There is plenty of evidence showing that the French and Haitian/Fongbe determiners do not have identical properties (see section 3.3.4.1). Whatever the precise characterisation of the parameter accounting for the availability of verb-doubling phenomena in particular grammars, it is clear from the data presented above that Haitian shares with its substratum languages the option that allows for these phenomena, whereas French has the opposite value for this parametric option. (For further discussion, see Lefebvre 1998a: 363-374.) From this perspective, it is hypothesised that verb-doubling phenomena became part of the Haitian Creole grammar in the following way: Speakers of the Kwa languages used the parametric value of their own grammar in setting the value for this parameter in the incipient creole. On the basis of the primary data

120

ISSUES IN THE STUDY OF PIDGIN AND CREOLE LANGUAGES

that they were exposed to, the first generation of Haitian native speakers identified the properties of the determiner in the language they were exposed to and deduced the availability of verb-doubling constructions. This is a reasonable hypothesis since, after 200 years of independent evolution, verb-doubling phenomena remain an important feature of the grammar of Haitian. The claim that this parameter setting was carried over into the creole by the substratum speakers is further reinforced by the fact that verb-doubling phenomena of the type described here are found only in those creoles which have substratum languages with this feature. For example, they are not found in the Pacific creoles. Verb-doubling phenomena thus do not constitute a general feature of creole languages; this constitutes a strong argument against the claim that all creole languages are alike (e.g. Bickerton 1984). Furthermore, given the rarity of verb-doubling phenomena among the languages of the world, the availability of such phenomena would be a marked option (e.g. Koopman 1986, for an extensive discussion of this issue). This conclusion runs counter to the claim by Bickerton (1984) and others who state that creole genesis involves setting the parametric options of UG for their unmarked values. 3.4.7 Summary As a result of the fact that syntactic clitics did not make their way into the creole, the value of the null subject parameter in the substratum grammar had to be reset. In all the other cases, the parametric values of the substratum grammars were transferred into the creole. As was shown throughout this section, this is largely due to the fact that the properties of the functional categories of the substratum lexicons have been reproduced in the creole through relexification. The absence of verb raising is linked to the absence of inflectional morphology. The availability of serial verbs is linked to the lack of derivational and inflectional morphology. The availability of the double-object construction is linked to the availability of the genitive case. The interpretation of negative quantifiers is linked to the availability of bare NPs. The availability of verb-doubling phenomena is hypothesised to be related to the properties of the determiner system. While the bulk of the parametric options of Haitian correspond to those of the substratum languages, none correspond to French parametric options. This fact strongly supports the view that the creators of a radical creole use the parametric values of their native languages in setting the parametric values of the creole. 3.5. Conclusion and consequences The data presented in sections 3.3 and 3.4 do support in a significant way the hypothesis of creole genesis outlined in section 3.2. They show that the bulk of the Creole's lexicon has been created by relexification; they also illustrate the fact

THE RELEXIFICATION ACCOUNT OF CREOLE GENESIS

121

that two other processes, reanalysis and levelling, play a role in the development of the creole. Lefebvre (1998a: 375-395) provides a detailed overall evaluation of the relexification hypothesis of creole genesis with respect to the lexicon, the semantic component and the parametric options. Several further questions for research pertaining to major category lexical entries, e.g. the Bantu component of early Haitian, the issue of the homogeneity of the substratum languages in creole genesis and the issue of dialect levelling, etc., are being addressed in a preliminary fashion. The data presented in this chapter call into question a number of assumptions in the field. For example, it has been claimed that a property that is shared by all contributing languages is more likely to enter the creole grammar than when the sources compete (see e.g. Singler 1988: 29). Some data discussed in this chapter constitute clear cases of the sources competing in different ways. For example, in section 3.3.1.3 on reflexivity, it was shown that the superstratum and substratum languages of Haitian do not have much in common in terms of how they encode reflexivity, and that furthermore, the substratum languages manifest variation as to how they encode reflexivity. Nonetheless, the idiosyncratic properties of the substratum languages have made their way into the creole. These data show that a Creole's source languages may contribute differential features in a principled way. Another example of data challenging shared assumptions in the field concerns the issue of markedness. Even since Bickerton (1984), it has been widely assumed that creole languages represent the unmarked case. Several subsets of data discussed in this chapter challenge this assumption. The verb-doubling phenomena discussed in section 3.4.6, the raising phenomena discussed in section 3.3.2.6, the availability of verb doubling phenomena discussed in section 3.4, etc. constitute examples in point. The data presented in this chapter strongly support the claim that relexification has played a major role in the formation of Haitian Creole. By hypothesis, this process plays a significant role in the formation of other creole languages as well. The fact that it can be demonstrated that relexification plays a role in the formation of various types of languages (e.g. mixed languages, pidgins, creoles) supplies evidence that this process is available to human cognition. It is a means of creating new languages (and new language varieties) in a relatively short time. The fact that it exists and the very nature of the process support Sproat's (1985) and Pranka's (1983) proposal that phonological representations are stored independently in the brain. The nature of relexification, the fact that it is available to human cognition and the fact that it is effectively used in the rapid creation of new languages have consequences for the theory of the transmission and acquisition of lexicons in situations where new languages, like creoles, are formed. Indeed, it is in the

122

ISSUES IN THE STUDY OF PIDGIN AND CREOLE LANGUAGES

nature of this process that lexical entries created in this way have phonological representations derived from phonetic strings in the lexifier language (thus showing discontinuity) but syntactic and semantic properties derived from the substratum language(s) (thus showing continuity). On the surface, then, it looks as if a totally new language has been created. In reality, however, the semantic and syntactic properties of the new lexicon are those of the substratum language lexicon(s). The properties of the original lexicon(s) are transmitted by adults and acquired by children even when the latter are presented with a relexified lexicon. Consequently, although situations where new languages are created by relexification involve a break in the transmission and acquisition of a language, there is no such break in the transmission and acquisition of semantic and grammatical properties (e.g. Lefebvre 1993b, 1996a, and section 3 of this chapter). As has been pointed out by Hopper and Traugott (1993: 211), the linguistic changes observed in the creation of pidgins and creoles "call into question the hypothesis that change occurs primarily in the transmission between generations, and is attibutable primarily to children." The very nature of relexification requires that those who apply it be adult native speakers in possession of a mature lexicon. Hence, the type of change resulting from relexification is initiated by adults. This constitutes a major piece of evidence against Bickerton's (1984) Language Bioprogram Hypothesis of creole genesis. The very fact that relexification exists as a cognitive process used to form new languages poses a problem for the genetic classification of the languages so formed. For example, Hall (1950: 203) classifies Haitian as a French dialect: "Haitian Creole is to be classified among the Romance languages, and especially among the northern group of the Gallo-Romance branch, on the basis of its systematic phonological, morphological, syntactical and lexical correspondences." Goodman (1964: 136) makes the following statement: "I do feel impelled to restate, however, that on the basis of no purely linguistic criteria for genetic relationship which have thus far been advanced, including that of 'parenté syntaxique' advanced by Sylvain (see 121-122), can Creole French be classified with any specific language other than French." The data presented here, however, cast considerable doubt on conclusions of this nature. Even though the phonological representations of Haitian Creole lexical entries can be associated with French phonetic strings, Haitian shares its lexical properties, morpho-syntax, concatenation principles and salient features of its parametric values with its substratum languages. Hence, it would appear that, from a typological point of view, Haitian Creole should be classified with its substratum languages. (For further discussion of the genetic classification of creole languages, see chapter 8). As a consequence of this state of affairs, scholars

THE RELEXIFICATION ACCOUNT OF CREOLE GENESIS

123

working on the reconstruction of language families should be aware that some languages (and not necessarily only those known as creoles) may have been created by relexification, thereby straying from the normal course of gradual linguistic change. The hypothesis of creole genesis presented here and supported by Haitian data also calls into question the assumption that all creole languages are alike, as is advocated by Bickerton (1981, 1984). To the best of my knowledge, this assumption was first challenged by Muysken (1988b) on the basis of a comparison of subsets of data drawn from various creole languages. In light of the hypothesis presented here, I would like to go one step further and claim that all radical creoles should show the division of properties between their source languages argued to exist in Haitian. Therefore, such creoles should have lexical entries with phonological representations derived from phonetic matrices of their superstratum language; the semantic and syntactic properties of these lexical entries, as well as the principles of concatenation and parametric values, should reproduce those of their substratum languages. While Pacific pidgins and creoles reproduce the specific features of the Austronesian languages (e.g. Keesing 1988), the Atlantic creoles reproduce those of their West African substratum languages. For example, while the pronominal system of Solomons Pidgin reproduces that of its Austronesian substratum languages in distinguishing singular, dual and plural, inclusive and exclusive first-person plural, etc (see Keesing 1988), the pronominal system of Haitian reproduces that of its West African substratum languages by not distinguishing first and second person plural (see section 3.3.1.2.1). Likewise, while the tense, mood and aspect system of Solomons Pidgin reproduces the details of the substratum languages, including preverbal and postverbal particles as well as a predicate marker (see Keesing 1988: 215; Sankoff 1991), the tense, mood and aspect system of Haitian reproduces the details of its substratum languages (see section 3.3.5.1). In a similar fashion, verb-doubling phenomena are only found in creoles that have a West African substrate. Systematic comparisons of other creole languages with their source languages should yield similar results.

CHAPTER 4

WHAT DO CREOLE STUDIES HAVE TO OFFER TO MAINSTREAM LINGUISTICS?* My first column is an essay on the relationship between creole studies and mainstream linguistics. I address the following questions: (1) In what sense do creole studies constitute a field? (2) What do creole studies have to offer to linguistics? (3) What progress has been made between 1994 and 1999? (4) What drawbacks have been overcome? and (5) What have creole studies contributed so far and what lies ahead? 4 . 1 . In what sense do creole studies constitute a field? There is a relatively large number of pidgin and creole languages that are spread across all continents (see Hancock 1971; Smith 1995). None of these languages are genetically related. Nevertheless, there is a field of creole studies. In what sense then do creole studies constitute a field? While this question is not a new one, it continues to be raised over and over again. On the basis of the available literature, I summarise the major proposals on which current debates build. One proposal holds that creoles represent the unmarked case (see e.g. Bickerton 1984; Seuren and Wekker 1986). This proposal has been the object of a long-standing debate that cannot be adequately summarised within the limits of this column. Suffice it to say here that this hypothesis is falsifiable. If subsystems of creole grammars can be shown to be marked, the hypothesis will be falsified. A fruitful discussion of this point, however, requires an independent measure of markedness (see e.g. Mufwene 1990, 1991; Muysken The content of this chapter was originally published as my first column for the Journal of Pidgin and Creole Languages (Lefebvre 2000a). I accepted to be the 2000-2001 columnist for JPCL in 1995 and I thank the editors for having given me this opportunity. This first column was dedicated to the memory of Chris Corne who tragically died in 1999. His work on Frenchbased creoles and on varieties of French in various geographical areas (e.g. Baker and Corne 1982; Corne 1977, 1979, 1986, 1995, 1999; etc.) constitutes a most important contribution to the field. Chris Corne's unique sense of humour and particular joie de vivre will be sorely missed by all those who knew him. I wish to thank the following people for their comments and suggestions on earlier drafts of this column: Anne-Marie Brousseau, Christine Jourdan, Robert Papen, Ingo Plag, Jean-François Prunet, Armin Schwegler, Jeff Siegel, Kevin Tuite, and Don Winford. Dany Adone, Philip Baker, Terry Crowley, Glenn Gilbert, John McWhorter, and Salikoko Mufwene provided specific information that was otherwise not available to me. Renée Lambert and Catherine Léger assisted me in doing the research on the references. Catherine Léger copy-edited the manuscript and Andrée Bélanger formatted it.

126

ISSUES IN THE STUDY OF PIDGIN AND CREOLE LANGUAGES

1981b). Some structures with a high index of markedness have been argued to be found in creole languages (see e.g. DeGraff 1992a, on the pro-drop parameter; Koopman 1986, on verb doubling phenomena; Muysken 1981b, on tense, mood, and aspect). This appears to invalidate the claim that creole languages represent the unmarked case [see also chapter 8]. Another proposal is that creoles are typologically similar in the sense that they share structural properties that set them apart from other natural languages (see e.g. Bickerton 1981, 1984; Seuren and Wekker 1986; etc.). Reduced morphology has long been proposed to be a shared feature of creoles (see e.g. Reinecke 1964). However, it has been noted that there are languages that have extremely reduced morphology, yet are not creoles. Vietnamese is a case in point. Thus, morphological richness/poverty cannot be viewed as a distinctive criterion. Furthermore, it has been pointed out on several occasions that the morphology of creoles is best evaluated on the basis of their substratum rather than on the basis of their lexifiers (see e.g. Alleyne 1966: 281; Brousseau, Filipovich, and Lefebvre 1989; Keesing 1988) [see also chapter 10]. It has also been observed that creole languages tend to be isolating (see e.g. Hagège 1985; Hesseling 1933; Mufwene 1986, 1990, 1991; Schuchardt 1979). Hence, for example, in creoles from different geographical areas, tense, mood, and aspect tend1 to be encoded by free morphemes occurring between the subject and the verb. Mufwene (1986, 1990) points out that this is also the case in creoles formed on the basis of almost exclusively agglutinative languages. Interestingly enough, in spite of this structural similarity among creoles, the semantic distinctions that are lexically encoded by tense, mood, and aspect markers in specific creoles reproduce those found in their respective substratum languages (see e.g. Keesing 1988; Sandeman 1998; Sankoff 1990, 1991, for Pacific creoles; and Bentolila 1971; Lefebvre 1996a, for Haitian Creole). Furthermore, the fact that creole languages tend to be isolating languages does not set them apart from other natural languages, for this characteristic is not solely a property of creole languages. For example, Chinese, not considered to be a creole, is also an isolating language. The claim that creole languages constitute a typologically discrete class is further weakened. On the one hand, scholars who have worked on a wide range of creole languages have shown that creoles are not as similar as they seem at first glance. A case in point is Muysken's (1988b) comparison of a sample of creoles from different geographical areas. This study shows that creoles differ fundamentally in a number of ways. On the other hand, comparative work on creoles and their source languages reveals that creoles tend to reproduce the 1 Note the use of the verb "tend". I do not exclude the possibility that there might be exceptions to this generalisation.

WHAT DO CREOLE STUDIES HAVE TO OFFER

127

syntactic and semantic properties of their substratum languages rather than those of their lexifier languages. Hence, Atlantic creoles tend to reproduce those of the West African languages (see e.g. Adam 1883; Alleyne 1980; Goodman 1964; Holm 1988; Huttar 1975; Lefebvre 1998a; Voorhoeve 1973), while Pacific creoles tend to reproduce those of the Austronesian languages (see e.g. Jourdan and Keesing 1997; Keesing 1988; Sankoff 1984, 1991). Thus, the features proposed in the past in support of the view that creole languages constitute a typologically discrete group of languages are not uncontroversial. I agree with Muysken's statement that "[t]he very notion of a 'creole' language from the linguistic point of view tends to disappear if one looks closely; what we have is just a language" (1988b: 16).2 I believe that one has to know the history of a particular language in order to identify it as a pidgin or a creole (see also Thomason 1997a on this point). In my view, what pidgin and creole languages share—regardless of where they are spoken or what their S-structure may be like—is the manner in which they are formed. This manner is what makes them a "legitimate" field of study (see Lefebvre 1998a). Mufwene (1997a) and Muysken (1988b) express similar opinions, though in different terms. The exact characterisation of this manner constitutes an unresolved debate in the field. McWhorter (1998a: 792-793, 797) has recently offered a new set of features which he claims come as a cluster in creole languages: (1) "minimal" usage of inflectional affixes; (2) "very little or no use of tone"; and (3) "generally" semantically transparent derivation. While it is not clear to me whether this proposal is formulated in terms that make it falsifiable (e.g. "minimal", "very little" or "generally"), it will undoubtedly give rise to a great deal of discussion. Forthcoming on this issue are the outcomes of the Chicago debate between McWhorter and DeGraff, held in January 2000, and a paper by Mufwene (2000) "Creolisation is a social, not a structural, process". It is to be hoped that McWhorter's proposal will also foster research on the morphology of a number of languages and on criteria used to identify this morphology as native to the creole. [For extensive discussion of this point, see chapter 10.] Another avenue of research suggested by the above proposal has to do with tones: what happens to tones in a situation where a creole develops from substratum languages with tones and a superstratum language without tones? 4 . 2 . What do creole studies have to offer to linguistics? As I have said above, creole languages are on a par with other natural languages when considered from the point of view of their grammar. Thus, the study of any grammatical subsystem of creoles contributes as much to linguistics as the 2

A similar view is also advocated in Mufwene (1994b).

128

ISSUES IN THE STUDY OF PIDGIN AND CREOLE LANGUAGES

study of any grammatical subsystem of other languages. I can see, however, several areas of linguistics where PCs may offer a unique perspective. The first one is the field of historical linguistics. I have suggested above that the link between creoles of different origins lies in the manner in which these languages are created. Assuming that this is a valid assumption, the central interest of creoles for linguistics is the problem of their emergence. Furthermore, there is a consensus in the literature that pidgins and creoles are created in a relatively short span of time compared with regular cases of linguistic change (see e.g. Alleyne 1966; Bickerton 1984; Chaudenson 1977, 1993; Ferraz 1983; Hall 1958; Hancock 1987; Hesseling 1933; Hymes 1971; Mintz 1971; Singler 1996; Thomason and Kaufman 1991; Van Name 1869-1870: 123, cited in Goodman 1964: 135; Voorhoeve 1973). Pidgin and creole languages thus constitute a unique opportunity to study cases of accelerated linguistic change (see e.g. also Plag 1994a on the question of linguistic change and creolisation). In recent literature, it has been proposed that adults are the principal agents of creole genesis (see e.g. Jourdan 1985; Lefebvre 1998a; Sankoff 1980 (ed.), 1991; Singler 1996: 199). As Hopper and Traugott put it, the linguistic changes observed in the creation of pidgins and creoles "call into question the hypothesis that change occurs primarily in the transmission between generations, and is attributable primarily to children" (1993: 211). Another related issue is the question of the genetic affiliation of creoles. There are four competing proposals in the literature. Because most of the phonological representations of creole lexical entries are derived from their lexifier language, some scholars classify them with their respective superstratum language. Because the syntactic and semantic properties of creole lexical entries are to some extent derived from their substratum languages, others classify them with their respective substratum. Since creoles tend to manifest the properties of their source languages in a principled way (see e.g. Huttar 1975; Keesing 1988; Lefebvre 1998a; Lefebvre and Lumsden 1989; Voorhoeve 1973)—let alone the innovations that they manifest—, other scholars consider them as mixed languages. On account of their mixed properties, some scholars even analyse them as genetically unaffiliated languages. The debate surrounding this issue is far from being resolved (see e.g. Hazaël-Massieux 1996; McWhorter 1998a, and the references cited therein). Given this situation, we would not want to exclude the possibility that other languages (e.g. languages not known to be creoles and difficult to classify due to their history) may have been formed in the same way creoles have. The second area of linguistics where pidgins and creoles may bring an important contribution is language contact studies. Pidgin and creole languages are mixed languages in that they derive their properties from their superstratum

WHAT DO CREOLE STUDIES HAVE TO OFFER

129

and substratum languages in a principled way (see the references above). Pidgin and creole languages thus instantiate extreme cases of contact between languages. As such, they intersect the field of language contact studies. The third area of linguistics where the contribution of pidgins and creoles may be special is the field of second language acquisition. Indeed, it has been proposed in some instances that pidgin and creole languages constitute particular cases of second language acquisition (see e.g. Alleyne 1971, 1980; Andersen 1980; Chaudenson 1993; Mufwene 1990; Schumann 1978; Siegel 1999a; Thomason and Kaufman 1991; Valdman 1980; some of the papers in Wekker (ed.) 1996; etc.). On the assumption that this is the case, there should be a link between the way pidgins and creoles are formed and the way second languages are acquired. Lumsden (1999a) proposes that this link lies in the process of relexification, which produces cases of transfer of a particular type. He argues that this process plays a role in both the formation of creoles and in some cases of second language acquisition as well. The fourth area of linguistics where pidgin and creole languages may offer a unique contribution is the field of sociolinguistics. Pidgins and creoles are generally created in multilingual communities (see e.g. Whinnom 1971) in need of a lingua franca (see e.g. Foley 1988; Hymes 1971; Thomason and Kaufman 1991). This particular situation gives rise to much variation between speakers of an early creole community (see e.g. Sankoff (ed.) 1980). Although this variation is not directly accessible in most cases, it can be observed in the further developments of a creole, including the contemporary stages of the language. This variation is undoubtedly a gold mine for the study of the nature of this variation (see e.g. Poplack and Tagliamonte 1996), the emergence and stabilisation of social patterns of variation (see e.g. Patrick 1999), and the mechanisms of levelling of this variation (see e.g. Siegel 1997)—both within a community and between communities. The fifth area of linguistics where pidgin and creole studies are relevant is applied linguistics. In most creole communities, the creole-speaking popula tion is educated in the lexifier language of this creole. However, phonologically similar lexical entries in the two languages (the creole and the superstratum) may not have the same syntactic and semantic properties (see e.g. Lefebvre and Lumsden 1994c). Comparative work on the lexicons and grammars of given creoles and their respective lexifier languages is needed in order to provide support to the educational enterprise. Furthermore, descriptive work on the lexicons and grammars of creoles is mandatory for communities that have chosen to conduct their literacy and formal education programmes in the creole. We may ask whether creole languages have something specific to contribute to theoretical linguistics. In Lefebvre (1998a), I suggest that the study

130

ISSUES IN THE STUDY OF PIDGIN AND CREOLE LANGUAGES

of creoles and their source languages may provide important insights into the structure of natural language lexicons and the components that define the parameters of variation between languages at both the syntactic and the phonological levels. Incipient creoles constitute principled compromises between their source languages (Lefebvre and Lumsden 1989). The nature of these compromises might teach us something about the form of natural languages. Although we are far from a breakthrough in this area of research, the discussions in DeGraff (1992b), Déprez (1992b, 1999), Koopman (1986), Lefebvre (1998a: 349-375), and Lumsden (1996) about how parameters are established in creole genesis point in this direction. In conclusion, the field of creole studies has plenty to offer to mainstream linguistics. 4 . 3 . What progress has been made between 1994 and 1999? Some of my predecessors as columnists have expressed mixed feelings with respect to the question of what creole studies have contributed so far to linguistics. In his paper entitled "The first seven years, fat or lean?", Muysken (1994a) sums up the contribution of various investigators to JPCL from the beginning of the journal until 1994. Major gaps are identified in the domain of phonology and in the field of creoles that are not European-based. Ironically, other crucial gaps identified by Muysken concern the relationship between creole studies on the one hand and historical linguistics and language contact studies on the other hand. Likewise, Winford points out the poor contribution of creole studies to sociolinguistics: 'The field of sociolinguistics has progressed a long way since the 1960s, but sadly, creole studies have generally failed to keep pace with it" (1997: 304). Is it not striking that crucial gaps have been identified exactly in these areas? On the basis of a sample of publications, I consider below what was done since 1994 and attempt to evaluate the progress. Table 4.1 shows the number of papers (articles, replies, rejoinders, and short notes) per topic that appeared in JPCL since 1994 up to the first issue of 1999.3 Table 4.1. Number of papers per topic in JPCL (1994-1999) 22 Syntax/semantics Phonology 5 0 Morphology Language change 2 Language contact 1 5 Sociolinguistics 4 External history Other topics 7 46 Total 3

The second issue of JPCL for 1999 was not out at the time I wrote this column.

WHAT DO CREOLE STUDIES HAVE TO OFFER

131

As can be seen in Table 4.1, papers on syntax/semantics outnumber those on other topics. There are five papers on phonology, partially filling the gap identified by Muysken (1994a). There are only two papers on language change and only one on language contact. The five papers on sociolinguistics partially fill the gap identified by Winford (1997). Four papers are dedicated to the external history of pidgin and creole communities. There are seven papers classified as "other topics", which do not fall into any of the above categories. Finally, seven papers classified under one of the topics in Table 4.1 also address the contribution of the substratum languages to a particular creole. The major gap revealed in Table 4.1 is in the area of morphology. Along with Arends, Muysken, and Smith (eds) (1995) and Spears and Winford (eds) (1997) which discuss various topics, books published by John Benjamins in the "Creole Language Library" collection since 1994 focus on the syntax and semantics of particular creoles (Clements 1996; Kihm 1994), on history (Arends (ed.) 1995), on language contact (Thomason (ed.) 1997), and on sociolinguistics (Escure 1997). Two other books published by John Benjamins that pertain to creoles (but are not part of the "Creole Language Library") are Adone (1994) on acquisition and Patrick (1999) on variation. These publications certainly fill some of the gaps identified above. Since 1994, a new series on creolistics has been created: the "West minster Creolistics Series", now published under the name of the "Battlebridge Publications". The distribution per topic of the papers in the first two volumes (Baker (ed.) 1995; Baker and Syea (eds) 1996) is shown in Table 4.2. Table 4.2. Number of papers per topic in the "Westminster Creolistics Series" (vols 1 and 2) Syntax/semantics 8 Phonology 1 Morphology 5 Language change 5 Language contact 4 Sociolinguistics 7 External history 3 Other topics 3 36 Total

Table 4.2 shows that, in addition to papers on syntax/semantics, there is a fair number of papers on language change, language contact, and sociolinguistics. The five papers on morphology fill a gap identified above in this area of research. Among the papers in Table 4.2, nine also discuss the contribution of the substratum languages to a particular creole. Within the same series, we also find an expanded and revised edition of Mühlhäusler's (1997) book on pidgin and creole linguistics (volume 3), a collection of papers by Baker and Bruyn (eds) (1999) on the significance of

132

ISSUES IN THE STUDY OF PIDGIN AND CREOLE LANGUAGES

early English Creole texts from St Kitts (volume 4), a book by Corne (1999) on varieties of French and French-based creoles around the world (volume 5), and a collection of papers by Huber and Parkvall (eds) (1999) on the possible role of diffusion in creole formation (volume 6). This information shows that the gaps identified by my predecessors have been at least partially filled, when research topics are considered. Furthermore, a comparison of Table 4.1 with Table 4.2 shows that the topics addressed in the "Westminster Creolistics Series" papers are more varied than those in JPCL. This suggests that there was indeed a need for another major series on pidgins and creoles. What is the situation of the research on particular pidgins and creoles? Table 4.3 presents the number of papers per language published in JPCL (1994-1999) and in the "Westminster Creolistics Series" (volumes 1 and 2). Only languages on which more than one paper was written are listed in Table 4.3. There are 18 languages in JPCL and 20 in the "Westminster Creolistics Series" that are the object of one paper only. Table 4.3.

Number of papers per language (discussed in more than one paper) JPCL "Westminster Creolistics Series" (1994-1999) (volumes 1 and 2) Haitian Creole 8 — Saramaccan 3 2 African American 3 — 3 Atlantic English-based creoles — Sranan 2 — Tayo 2 2 Trinidad and Tobago creoles 2 — 2 Hawaiian pidgins and creoles — Melanesian Pidgin 2 — Mauritian Creole 3 — 2 Krio — Isle de France Creole 3 — 2 Australian Pidgin English — 5 Not applicable — Total 30 16

The languages listed in Table 4.3 mainly have French or English as a lexifier language. In addition, two books in the "Creole Language Library" (John Benjamins) are on Portuguese-based creoles (Clements 1996; Kihm 1994). Progress in the study of Spanish-based varieties is extensively discussed in Schwegler (2002). On creoles that are not European-based, the following publications are available: on Sango (Morrill 1997; Pasch 1997; Samarin 1994, 1997; Thornell 1995, 1997; Walker and Samarin 1997; and Zribi-Hertz and Bingaba-Ngaima 1994); on Kituba (Mufwene 1994a, 1997b); on Arabic-based

WHAT DO CREOLE STUDIES HAVE TO OFFER

133

pidgins and creoles (Owens 1997). These publications show that there has been some progress in the study of creoles that are not European-based. The languages in Table 4.3 are spoken in Africa, the Caribbean, the Americas, Hawaii, New Caledonia, Australia, Mauritius, the Seychelles, and the Pacific.4 Thus, the main geographical areas where creoles are spoken are represented in the sample. Although the above survey provides a partial picture of recent publications in the field, it shows that a wide range of topics, languages, and geographical areas are covered in recent studies. The major gaps identified for the period prior to 1994 have been at least partially filled between 1994 and 1999. Studies in phonology and morphology are still called for, as well as studies on creoles that are not European-based. All in all, however, we can conclude that there has been progress. The visibility of creole studies has also increased since 1990. This can be observed on the basis of publications on pidgin and creole languages in mainstream linguistics journals. For example, while there are no articles on pidgins or creoles published in Language between 1989 and 1997, five have appeared in 1998 and 19995: Goodman (1999), McWhorter (1998a), Roberts (1998, 1999), and Romaine (1999). An entire issue of Linguistics, edited by Seuren and Mufwene (1990), is dedicated to creole linguistics, and since then, two papers on creoles have appeared in the journal: Seuren (1995) and Syea (1997). 4 . 4 . What drawbacks have been overcome? 4.4.1 The Tower of Babel It is quite tempting to look at creole studies as some sort of Tower of Babel. Commenting on the lack of consensus in the field, Winford makes the following remark: 'The lack of consensus is not confined to the creole fraternity, but extends to our relationship with noncreolists as well" (1996: 71). In order to illustrate this point, Winford shows that the vocabulary used in papers on tense, mood, and aspect in creole studies differs not only between authors, but also from the more standard vocabulary used in mainstream linguistics. This has the double effect of making comparative work between creole TMA systems quite difficult and making access to work on creole TMA systems quite impossible for non-creolists. Winford concludes: "[...] we must stay in tune with scholars in 4

Pacific creoles are often discussed in other sources than those mentioned here. For example, for the period 1994-1999, the Linguistics and Language Behavior Abstracts (LLBA) lists 13 entries for Tok Pisin, 5 for Bislama, and 7 for Hawaiian. See also the references in Wurm, Mühlhäusler, and Tryon (eds) (1996). 5 At the time I wrote this column, there were only three issues of Language available for 1999.

134

ISSUES IN THE STUDY OF PIDGIN AND CREOLE LANGUAGES

the fields of semantics and language typology who have contributed so much to our understanding of TMA systems in the world's languages. At the very least, we should bring our terminology in line with theirs, unless we have really compelling reasons for doing otherwise" (1996: 82). I could not agree more with Winford's conclusion. The very subject he chose for his column "Common ground and creole TMA" is a good example. Moreover, the practice of developing a rather sophisticated inner-group vocabulary extends to discussions pertaining to other dimensions of creole studies. Another example, also discussed in Mufwene (1997a), is the range of terms used to refer to language types in the field: in addition to "jargons", "pidgins", "creoles", and "koines" (for which there is no shared definition as yet, but see Thomason 1997a), we find the expressions "stable pidgins", "expanded pidgins", "creolised pidgins", "radical creoles", "decreolised creoles", "semi-creoles", "restructured varieties", etc. Even for those who are familiar with the literature in the field, it is not always clear how the terms are used in specific instances. Since these terms all refer to types of language varieties and since more types are being found, it is not clear whether the list of terms will ever constitute a closed set. Furthermore, these types of language varieties constitute outputs to linguistic processes. In the area of processes as well, a wide range of terms are used. Along with "pidginisation", "stabilisation", "nativisation", "creolisation", and "decreolisation", we find "simplification" and "expansion". Although these terms do not specifically refer to linguistic processes, they are often used as such in creole studies. Pidgins and creoles are created in order to provide speakers of different lexicons with a common vocabulary so as to facilitate communication. Is it not ironic that the vocabulary developed to talk about these languages and the processes at work in their formation has the effect of hindering communication between scholars? 4.4.2 Isolating Atlantic creolesfrom Pacific creoles In the same vein, I think the field would gain a great deal if the languages that it studies were not subdivided ad absurdum. For example, why consider Tok Pisin apart in discussing creole languages or treat Atlantic creoles and Pacific creoles separately, as is suggested in some instances. The field will not gain anything by making each case a particular one; it will gain a great deal, however, by considering the largest possible number of cases as well as the similarities and differences between them. A welcome effort in this direction is the collection of papers by Byrne and Holm (eds) (1993) on Atlantic and Pacific creoles. Much more work needs to be done in this direction.

WHAT DO CREOLE STUDIES HAVE TO OFFER

135

4.4.3 Isolating the study of pidgin and creole languages from that of mixed languages Pidgins and creoles are mixed languages, but their study has traditionally been kept apart from that of mixed languages. Is it not surprising that the first collection bringing together case studies of pidgins, creoles, and mixed languages dates from only a few years ago (Thomason (ed.) 1997). In my view, this is particularly surprising for two reasons. First, several studies on mixed languages were done in the last twenty years (see e.g. Bakker and Mous (eds) 1994; Muysken 1981a). Second, the first attempt to link the emergence of these languages through some [cognitive] process goes back to Muysken (1981a). Between 1981 and 1997, creoles and mixed languages have been kept apart. This is regrettable since the most promising idea that I have seen expressed concerning the relationship between creoles and mixed languages is that of Muysken (1981a: 77). At the end of his Media Lengua paper, in which he argues that relexification was the main process at work in the formation of this mixed language, he writes: "If it is the case that the Caribbean creoles show numerous African survivals in their syntax and semantics, then I think we can argue that it is not interference which led to these survivals, but relexification" (1981a: 77). This remark has long been a source of inspiration for my own work. Although Muysken later amended his earlier claim by restricting relexification to cases of mixed languages (see Muysken and Smith 1990: 884), I am convinced that his original statement pointed in the right direction. Clearly the components of the situations in which mixed and creole languages emerge differ in some points (see e.g. Bakker and Mous (eds) 1994)—namely, whether second language acquisition is implied or not, the number of languages involved, and the reasons why these new languages are created—. I believe, however, that relexification is the main process at work in both cases (in spite of the fact that the overall outputs of the process may not be the same in both cases). I hope that my discussion of this point in Lefebvre (1998a: 19-30) will help build a bridge between the study of the genesis of mixed languages and that of creole languages. 4.4.4 Setting issues outside the research paradigms of subdisciplines of linguistics Another shortcoming is that research [in PC studies] is often framed outside the paradigms of subdisciplines of linguistics. A case in point is the debate on creole genesis. Indeed, this debate has been set in terms of universalist (e.g. Bickerton 1984), superstratist (e.g. Chaudenson 1992), and substratist (e.g. Alleyne 1980) approaches. The universalist approach rests on the assumption that creole languages are typologically similar and on the claim that these languages manifest the unmarked options of universal grammar (see the discussion above).

136

ISSUES IN THE STUDY OF PIDGIN AND CREOLE LANGUAGES

The superstratist approach rests on the similarity between the forms of the lexical entries in a creole and those in its lexifier language; this approach is challenged by Mufwene in the following terms: "One of the problems with the superstrate hypothesis is the absence of any explanation for why creoles lexified by European languages do not correspond to any particular dialect of their lexifiers" (1996b: 166). The substratist approach rests on the syntactic and semantic similarities between creole languages and their substratum languages. These similarities are sometimes hidden due to the fact that, as has been pointed out by Hall (1958), creoles have retained few, if any, visible features (phonological forms) of their substratum languages. None of these three approaches can account for all the properties of creoles. The universalist approach cannot account for the differences observed between creoles nor for the marked options that they manifest (as discussed above). None of the three approaches can account for the fact that all creoles manifest properties of both their superstratum and substratum languages in a principled way (see Lefebvre and Lumsden 1989 and chapter 2 for further discussion of this point). In addition to the fact that these approaches cannot account for the properties of pidgin and creole languages, they set the debate on creole genesis outside the paradigm of historical linguistics. In my view, this has the consequence of keeping the debate on creole genesis outside mainstream discussions on language change in general. In the last decade, there has been progress with respect to this point. The problem of creole genesis has been cast within the framework of the processes otherwise known to play a role in language genesis and language change in general: relexification (see e.g. Lefebvre 1998a; Lefebvre and Lumsden 1989; Lumsden 1999a), reanalysis (see e.g. Bruyn 1996; Lefebvre 1996a; Plag 1993; Romaine 1999; Sankoff 1990, 1991), and levelling (see e.g. Harris 1991; Lefebvre 1998a; Lumsden and Lefebvre 1994; Mufwene 1990; Mühlhäusler 1980; Siegel 1997). My evaluation of this methodology, in Lefebvre (1998a), is that, with these three processes—and a theory on how they interact—, it is possible to account for the bulk of the properties of creole languages. Although there is still no consensus on the problem of the genesis of creoles, I believe that we have much to gain in keeping the framework of discussion on the topic within the terms of historical linguistics. 4.5.

What have creole studies contributed so far and what lies ahead? I begin with contributions to theoretical linguistics. When the bioprogram hypothesis was in vogue, creole studies were at the cutting edge of theoretical linguistics and Bickerton's work is still cited in relation to the innateness hypothesis. [Now that this hypothesis has been falsified, see chapters 3, 9 and 10, and the references therein], an important question concerns what creole

WHAT DO CREOLE STUDIES HAVE TO OFFER

137

studies now have to offer to mainstream linguistic theory. A promising area is that of the parameters of variation, both in phonology and syntax. Such contribution is conditional to the availability of three-way comparisons of particular creoles and their source languages. As more scholars are shifting their object of study towards such three-way comparisons (see e.g. Migge 1998a), the contribution of creole studies to the theory of parametric variation promises to be extremely fruitful. The last decade has witnessed the emergence of research on earlier stages of creoles based on written sources. Building on pioneer work by Arends (1989) and Smith (1987), papers written within the above framework are published in Arends (ed.) (1995), Baker and Bruyn (eds) (1999), Huber and Parkvall (eds) (1999). This type of research has added significant insights to our understanding of creole genesis and development. For example, a number of proposed grammaticalisation scenarios for certain structures crumble down when diachronic facts are considered (see e.g. Bruyn 1996; Plag 2002). Diachronic research as documented in Baker and Bruyn (eds) (1999) substantiates claims that the English-based Caribbean creoles used to be more similar to each other than they are now. This has consequences for the origins debate. Work on written sources has thus proved to be most relevant to a number of issues and more work along these lines is certainly welcome. However, this methodology raises the double problem of the availability and reliability of written sources. Such written sources are not available for all creoles. Furthermore, the reliability, in terms of their representativity, of these written sources has to be assessed. A discussion of this problem for Haitian Creole written sources can be found in Lefebvre (1998a: 68-70). The debate on creole genesis has been quite significant in the last decade. Several books are dedicated to this topic and offer competing views (see e.g. the papers in DeGraff (ed.) 1999; Hazaël-Massieux 1996; Keesing 1988; Lefebvre 1998a; McWhorter 1997b; etc.). The debate on diffusion, which goes back to the sixties, has also been revived by a number of authors (see e.g. Goodman 1985, 1999; the papers in Huber and Parkvall (eds) 1999; McWhorter 1997a; Roberts 1998, 1999). If anything, the above research shows that it is no longer possible to address the question of creole genesis on the basis of the lexifier languages alone, and that the contribution of the substratum languages has to be taken into account as well. No matter which account of creole genesis will eventually be adopted, we need to explain why some substrate features do not make it into the creoles. A likely filter is the superstratum language itself (see e.g. Lefebvre 1998a; Lefebvre and Lumsden 1994c; Siegel 1999a: 36-37). This remains to be sufficiently documented. We also need an auxiliary theory of data processing which will explain how the superstrate data are processed by

138

ISSUES IN THE STUDY OF PIDGIN AND CREOLE LANGUAGES

speakers of the substratum languages in creole genesis (see Plag 2000). Hopefully, the formulation of such a theory will connect creole studies to other subdisciplines of linguistics concerned with data processing in acquisition. The topic of levelling is a rather new one in the literature on pidgins and creoles (see Lefebvre 2001a; Siegel 1997). Studies on communities where incipient creoles develop will be most important to our understanding of how this social process applies to early creole communities (Escure 1997; Jourdan 1985; Patrick 1999). Creole languages have long been perceived as non-legitimate languages. Persistent work by linguists has shown, however, that these languages are indeed natural languages. There are now several creole studies programmes in universities. There are also dictionaries and descriptive grammars of several creole languages. Furthermore, some creoles are now being taught as second languages. Educational problems concerning the status of creoles and black vernaculars have recently been raised publicly. In this respect, the debate on Ebonics has certainly marked the last decade (see e.g. Baugh 1999; McWhorter 1998b; Perry and Delpit (eds) 1998; Rickford and Rickford 1995; Rickford and Rickford to appear). There is much progress to be made in the area of applied linguistics (for directions in this area, see Rickford 1997; Siegel 1999b). What have creole studies contributed so far? My own assessment is the following. Although there has been much original research—several major publications in the field, and more visibility of the research in the field—the contribution of creole studies to mainstream linguistics has not been more important than that of other languages. On the one hand, this assessment should not come as a surprise given the position taken in section 4.1 that, from a structural point of view, creole languages are no different from other natural languages. On the other hand, this assessment should not be considered as too positive, in view of the fact that there are specific areas of linguistics where pidgins and creoles have been identified as potentially offering a unique perspective (see section 4.2). So, in my view, genuine contributions of creole studies to mainstream linguistics will come from break-throughs on topics identified in the section entitled "What do creole studies have to offer to lingusitics?" What more lies ahead? The proceedings of two panels, edited by G. Gilbert (2002)—which were part of the 1998 Society for Pidgin and Creole Linguistics meeting (Linguistic Society of America, in New York)—propose new directions for research on all the topics raised in this column. In my view, what the field needs most of all, however, is new databases that will feed the debates on all the issues under discussion.

CHAPTER 5

ON DATA* In the introduction to his book on Fula, De Guiraudon (1894: V-VIII) reviews literature on the language. The following passages are representative of his style of presentation. ANONYME.—Vocabulaire foule, publié en 1845, dans le tome II des Mémoires de la Société ethnologique, d'après un manuscrit de la Bibliothèque nationale. Ce vocabulaire remonte au siècle dernier, et semble avoir été recueilli par un missionnaire français : outre de nombreuses erreurs, l'écriture du manuscrit est très mauvaise, de sorte que l'éditeur, ne connaissant pas la langue, a mal lu presque partout et a ajouté de nouvelles erreurs (1894: VI). [ANONYMOUS.—Vocabulaire foule, published in 1845, in volume II of the Mémoires de la Société ethnologique, based on a manuscript of the Bibliothèque nationale (National Library). This vocabulary dates back to the last century and seems to have been gathered by a French missionary. The manuscript comprises numerous errors; moreover, the quality of the writing is very poor. Thus, the editor, who did not master the language, misread almost everywhere and added errors of his own.] REV. C.A.L. REICHARDT.—Grammar of the Fulde language (London 1876). L'auteur de ce long et ennuyeux livre, mercenaire allemand au service des missions anglaises, ne semble pas avoir cherché à apprendre la langue de la bouche des indigènes : la chose était sans doute au-dessous de sa dignité, et il a une manière bien plus originale de procéder. Au lieu de déduire les principes de la grammaire de l'étude des textes recueillis par son prédécesseur et homonyme, il a préféré inventer d'emblée des théories grammaticales fausses et absurdes, après quoi il n'a pas craint de falsifier les textes pour les mettre d'accord avec les règles écloses dans son cerveau malade (1894: VI). [REV. C.A.L. REICHARDT.—Grammar of the Fulde language (London 1876). The author of this long and boring book, a German mercenary at the service of the English missions, does not seem to have tried to learn the language through native speakers: this task was probably beneath his dignity and his way of proceeding was much more original. Instead of deducing the grammar principles from the study of the * The content of this chapter was originally published as my second column for the Journal of Pidgin and Creole Languages (Lefebvre 2000b). The research underlying this column was financed by SSHRCC, FCAR, and UQAM's institutional research funds. I would like to thank the following people for their very useful questions and comments on an earlier version of this column: Anne-Marie Brousseau, Viviane Déprez, Robert Papen, Ingo Plag, Jean-François Prunet, Armin Schwegler, Jeff Siegel, and Donald Winford. The usual disclaimer applies. I would also like to thank Renée Lambert for helping me with the documentation, Catherine Léger for copy-editing the manuscript, and Andrée Bélanger for formatting it.

140

ISSUES IN THE STUDY OF PIDGIN AND CREOLE LANGUAGES texts gathered by his predecessor and namesake, he preferred to invent false and absurd linguistic theories straightaway; afterwards, he did not mind altering the texts so as to make sure that they would be in agreement with the rules hatched in his sick mind.]

In the linguistic literature of the late nineteenth century, it is not rare to find these kind of remarks on data. While the tone of the debate surrounding this topic may have improved in some cases over the last century, there are still a lot of ongoing discussions on data on the basis of languages of various families. In recent literature in creole studies, several facets of the problem surrounding linguistic data have been at the heart of controversies. For example, some authors have raised the problem of what constitutes a linguistic fact in creole studies (e.g. Bickerton 1996; DeGraff 1999). Others have raised the question of whether historical texts constitute a valid source of data for linguistic analysis (e.g. Muysken 1995). The problem of inconsistencies in some types of linguistic data has also been addressed (e.g. Labov 1975). This column is a discussion on data, based on some thirty years of experience in data collection and analysis, mainly carried out on Quechua, French, Martinican Creole, Haitian Creole, and Fongbe. My thoughts are organised around three major themes: the non-neutral character of linguistic data, how to overcome the limits of particular databases, and the problem of "inconsistencies" in elicited data. 5 . 1 . The non-neutral character of linguistic data Data collection is not a neutral activity. It is always designed within the framework of a particular methodology. Different research paradigms and questions call for different methodologies, hence for different types of databases. Data are thus not independent from research paradigms and questions. For example, the content of lexicons (or dictionaries) varies according to the research tradition of the authors. The methodology used to establish official terminologies (e.g. Chansou 1997; Gacic 1994; Nagao 1994; Thoiron et al. 1997) differs from that used to record the lexical entries of everyday speech. For one thing, whereas the latter may provide all the forms used to refer to a given object, including dialectal or regional variants, the former will provide the one form that has been selected as "official". Likewise, the methodology established by Weinreich (1984) for lexicographers specifies that the definitions of lexical entries should be valid for the entire linguistic community rather than for particular idiolects. In contrast, lexicons built in the tradition of lexical semantics are based on the mental lexicons of individuals. Thus, in this approach, the content of lexical entries corresponds to particular idiolects. In writing lexicons (or dictionaries), lexicographers and lexical semanticists have different goals. Lexicographers seek to make dictionaries of particular languages. Lexical semanticists, however, seek to describe the knowledge

ON DATA

141

speakers have that enables them to use particular lexical entries. The methodology that guides the content of lexical entries is thus not exactly the same in the two traditions. For example, in the English-Haitian Creole lexicon in Valdman (1996), we find that the English verb 'to escape' may be rendered by two Haitian words: chape and sove. In Valdman (1996), the content of these two lexical entries is as in (1). (1)

a. b.

chape escape, make it, pull through, rescue sove break out, escape, fly the coop, get out sove ak/avè(k) make off with sove kite flee

HAITIAN HAITIAN

The English lexical entry 'to escape' described by Levin (1989: 120), a lexical semanticist, is as in (2). (2)

a. b. c.

d. e.

f. g.

The convict escaped. Preposition Drop: a) The convict escaped from the police. b) The convict escaped the police. * Causative/Inchoative Alternation: a) The convict escaped. b) *The collaborators escaped the convict. (on the reading of "cause to escape") Depictive but not resultative phrases: The convict escaped exhausted. Adjectival Passive/Perfect Participles: an escaped convict, the recently departed guest (a convict that has escaped, not a convict that someone has escaped from) *an escaped jail *The convict escaped the soles off his shoes. *The convict escaped his way to freedom.

The content of the lexical entries in (1) and (2) reflects the different goals of lexicographers and lexical semanticists, respectively. This shows that the type of data researchers collect and the type of information recorded in lexical entries are not independent from research paradigms and questions. Likewise, we can contrast the types of grammars that can be written on a given language. For the present purpose, I will contrast prescriptive grammars with generative grammars. Prescriptive grammars seek to provide the official or accepted way of saying things in a given language. They thus provide a list of licit structures and sometimes a list of structures identified as "do not say". The structures identified as "do not say" are often possible ones in the language, in the sense that some people use them. They are, however, identified as illicit on

142

ISSUES IN THE STUDY OF PIDGIN AND CREOLE LANGUAGES

the basis of normative judgements. In contrast, generative grammars seek to provide an account of the knowledge speakers have that enables them to use their language. The databases of generative linguists mainly consist of grammaticality judgements. Generative grammars thus provide the structures that are judged to be grammatical by a given speaker; they mark as ungrammatical those that have been identified as such by the same speaker. Structures considered illicit by normative grammarians are not necessarily illicit for generative linguists. "Truly" ungrammatical sentences (in the generative grammar sense) are usually not provided by normative grammarians; they are, however, crucial to the analyses produced by generative linguists. The two types of grammars thus call for databases that overlap in some points, but that are far from being identical. Another contrast between two research paradigms further shows that data collection is not a neutral activity. The methodology guiding the study of ongoing linguistic change within a given community, as established by Labov and his associates, calls for large corpora of natural speech data drawn from a representative sample of members of this community (e.g. Labov 1972a; Sankoff (ed.) 1980; etc.). Similarly, longitudinal studies of child language acquisition require corpora of recorded data drawn from the speech of children (e.g. McDaniel et al (eds) 1996, and the references therein). In contrast, the methodology of what is being referred to in the literature as "field linguistics" does not call for the same type of data. The aim of field linguistics is to provide descriptions of the languages of the world in order to reveal the range of structures available to human languages. These descriptions can be used to build a database useful for comparing the languages of the world. In field linguistics, the informant is the source of information and the evaluator of the utterances submitted by the investigator (e.g. Bouquiaux and Thomas 1992; Burling 1984; Kibrik 1977; Longacre 1964; Payne 1997; Samarin 1967; Vaux and Cooper 1999; etc.). Samarin stresses the fact that "(...) without an informant one cannot test hypotheses (...) and cannot make statements concerning the productivity of morphemic relations—one cannot predict" (1967: 23). Bloomfield was a leader in this tradition of research, mainly through his work on Algonquian languages (e.g. A Leonard Bloomfield Anthology, edited by Hockett 1970, and "Leonard Bloomfield's descriptive and comparative studies of Algonquian" by Goddard 1987). Most of the books which are part of grammar collections of several major publishing houses are based on this tradition of research. For example, in March 2000, Mouton de Gruyter had in one of its collections twenty grammars of languages as diverse as Pipil, spoken in El Salvador (Campbell 1985), Breton (Press 1986), Kilivila, spoken in the Trobriand Islands (Senft 1986), Slave, an Athapaskan language (Rice 1989), Afrikaans, spoken in South Africa (Donaldson 1993), Berbice Dutch Creole (Kouwenberg 1994), Wardaman,

ON DATA

143

spoken in the Northern Territory of Australia (Merlan 1994), Tukang Besi, spoken in Indonesia (Donohue 1999b), etc. The grammars of a significant number of languages (e.g. African, Amerindian, Asian, Australian, European, Indonesian languages, etc.) have been described within the research tradition of field linguistics. It goes without saying that the methodology and the type of data that define this paradigm of research cannot be used to address the questions of researchers working in sociolinguistics and language acquisition. The reverse is also true. The large corpora of data required in sociolinguistics and language acquisition could not be used as such in the research programme set forth by field linguists. This is another example showing that the type of data researchers collect is determined by their research paradigms and questions. All the examples above show the non-neutral character of linguistic data. Several authors have even pointed out that linguistic theories constitute instruments that are helpful in searching for data relevant to solving problems (e.g. Aoun 1992; Botha 1976; Mulder 1996). As Aoun puts it: "(...) the uncovering of the data is theory driven: it is the theoretical or analytical apparatus that explicitly or implicitly guides linguists in their search for the data" (1992: 77). Furthermore, as was illustrated above, what should be recorded in the course of data collection for a given project depends on the nature of individual research programmes. It should be pointed out that the non-neutral character of data has become a central notion in epistemology. As advocated by Popper (1972), researchers need to know what to observe and why in order for their observations to be meaningful. This is true for all sciences, including linguistics. In this respect, we may wonder why the types of data produced by the methodologies of field linguistics and generative linguistics have been so unpopular with many colleagues in creole studies, given the fact that these methodologies are being used successfully for a wide range of languages. Since scholars who work on creole languages come from different research traditions, it should not come as a surprise that the data they report on are of different types. In my view, data coming from different research traditions on a given language are complementary. They reflect the different angles from which the linguistic facts of a given language have been addressed. Each type of data contributes something to our overall understanding. Each type presents advantages and disavantages. There are ways of overcoming the limits of individual types of databases, however. These topics are discussed in the next section. 5 . 2 . How to overcome the limits of particular types of databases There are different types of data that researchers can use as sources for their analyses. The following paragraphs discuss the advantages and disadvantages of

144

ISSUES IN THE STUDY OF PIDGIN AND CREOLE LANGUAGES

the various types of databases and present ways of overcoming their respective limits. Older texts in a given language constitute a possible source of data for linguistic analysis. This source of data may provide information on earlier stages of a particular language. As has been pointed out by Muysken, however, "part of the problem in dealing with historical materials is determining exactly what type of speech is reflected by them. This problem is even greater in creole societies, where often dramatic linguistic differences occur within one speech community" (1995: 335). This observation by Muysken, based on Negerhollands historical materials, carries over to Haitian historical materials. As has been pointed out in Lefebvre (1998a: 68-69), the Haitian written sources covering the period between 1776 and 1936, listed in Baker and Corne (1982: 273-274), have been questioned with respect to whether they reflect the speech of the majority of the African slave population in Haiti at the time they were written. Chaudenson (1977: 259), Dejean (cited in DeGraff 1993a: 90, note 56), and Lefebvre and Lumsden (1994a: 56) are of the opinion that they do not. Hence, for these authors, Haitian historical materials are not considered as a valid source of data (but see Carden and Stewart 1988: 26-27, for a more moderate position on this matter). Whether historical materials constitute a valid source of data for linguistic analysis may vary from language to language, depending on the adequacy of the written sources. Hence, while they may sometimes be considered as a valid source of data (e.g. Muysken 1995, on Negerhollands), this is not always the case (e.g. Lefebvre 1998a: 68-69, on Haitian). Large corpora of recorded and transcribed material present advantages and limits (e.g. Isquerdo 1998; Issoufi 1998; Kuo 1997). As for the advantages, they allow us to make word and structure frequency analyses, probabilistic analyses, discourse analyses, and language variation analyses along various dimensions. As for the drawbacks of recorded texts, we find the following: recorded texts do not contain readily available information on the meaning of words and structures any more than written texts do. For example, the lexical entries sa and sila were shown to manifest slightly different interpretive patterns across Haitian speakers. The three patterns of grammar reported on so far are listed in (3) (from Lefebvre 1999b) [see also chapters 3 and 9]. (3)

G1 G2 G3

sa [+proximate] sa [α proximate] sa [α proximate]

sila [-proximate] sila [-proximate] sila [α proximate]

HAITIAN

ON DATA

145

The first pattern, identified as G1, is attested in Goodman (1964: 51) and Tinelli (1970: 28). The second pattern, identified as G 2 , is attested in Étienne (1974: 154), Lefebvre (1997), and Sylvain (1936, as discussed in Lefebvre 2001a). The third pattern, identified as G3, is attested in Férère (1974: 103), Joseph (1988: 112), Valdman (1978: 194), and Valdman et al (1981). Texts do not readily provide this type of information. Likewise, in the predicate cleft construction, the clefted constituent contains only a "copy" of the verb. The clefted constituent may, however, be assigned several contrastive interpretations that go beyond the "copy" of the verb. Consider the three contrastive interpretations in (4). (4)

Se monje Jan manje pen an. HAITIAN it.is eat John eat bread DEF a. 'It is EATING the bread that John did.' (he did not throw it away) b. Tt is EATING THE BREAD that John did.' (he did not wash the dishes) c. Tt is eating THE BREAD that John did.' (he did not eat the apple) (=(44) in Lefebvre 1990)

While texts may contain occurrences of the predicate cleft construction, they do not provide the linguist with the possible interpretations of this construction. The only access to the meanings of this type of structure is through work with native speakers. In languages like Haitian and Fongbe (e.g. Avolonto 1992; Lefebvre 1996a; Lumsden 1995a; etc.), the temporal and aspectual interpretations of clauses depend on the aspectual properties of the situation (or aspectual classes) described by the clause. Texts do not provide these interpretations. Again the discovery of the pertinent interpretive facts is made possible through discussions with native speakers on the meaning(s) of clauses. Another disadvantage of recorded corpora is that, while they present positive evidence attesting the existence of a specific structure in a given language, they fail to present negative evidence. That is, they do not present impossible or ungrammatical structures. As was mentioned above, ungrammatical data are crucial to some research paradigms. Again, this type of data can only be provided by means of elicitation sessions with native speakers. Elicited data, consisting of grammaticality judgements by individual speakers, also have their advantages and limits. As for the advantages, they allow us to establish impossible as well as possible structures in the various subsystems of a given grammar (e.g. Aoun 1992; Botha 1976; Schutze 1996). Furthermore, as was mentioned above, this type of data is the only one giving access to the meaning of words and structures for particular speakers. As for the limits, we find the problem of inconsistencies (within a given speaker and between speakers), a problem that will be taken up in the next section.

146

ISSUES IN THE STUDY OF PIDGIN AND CREOLE LANGUAGES

Given the fact that each type of data presents advantages and limits, the entreprise of establishing data on a given language should be considered a collective one, whereby scholars from different research paradigms report on what they find. In my view, this is how, with time, data on a given language get established. The distribution of ki in Haitian Creole is one example of this process. In the literature on Haitian Creole prior to 1980, there is very little information on subject extraction in Wh-questions. For example, Sylvain (1936: 69) provides no examples of subject extraction. Hall (1953), Pompilus (1976), and Valdman (1970, 1978) each provide one or two examples of subject extraction involving short-distance Wh-movement, as is illustrated in (5). (5)

Ki-moun ki rele rel sa-al HAITIAN which-person RES shout shout DEM-DEF 'Who shouted that shout?' (from Pompilus 1976: 166)

As is pointed out by Valdman (1970: 204), when the subject has been questioned, the morpheme ki must appear in the vicinity of the verb.1 Obligatoriness of ki in sentences of the type of that in (5) is shown by the ungrammaticality of the sentence in (6), where ki is missing. (6)

*ki-moun which-person

_ _

te vini ANT come

an HAITIAN DEF (=(16) in Koopman 1982a: 211)

To my knowledge, Koopman (1982a) is the first author to have provided examples of long-distance Wh-movement involving the subject. Such a case is illustrated in (7). (7)

Ki-moun Jan kwè Mariz te di ki te vini an? which-person John believe Mariz ANT say COMP ANT come DEF 'Who is it that John believed that Mariz said that came?' (=(34) in Koopman 1982a: 216)

Koopman (1982a: 219-220) points out that ki must occur in the embedded clause from which the subject has been extracted by long-distance Whmovement. Without the embedded ki, the clause in (7) would be ungrammatical. Furthermore, Koopman shows that ki only occurs in cases of Wh-subject extraction, whether it involves short- or long-distance movement. It cannot occur in other cases; namely, it is excluded in cases of object or adjunct extraction. 1 There are competing analyses for ki in the literature. Some authors analyse it as a complementiser governing the empty subject position; others, as a resumptive pronoun occurring in subject position. The glosses in the examples respect the authors' analyses [see chapter 3].

ON DATA

147

Déprez (1992a, 1992b, 1994), Law (1992, 1994b), Lumsden (1990), and Sterlin (1988) provide data that show the same distribution for ki as the one in Koopman.2 DeGraff reports on a slightly different distribution of ki, whereby ki "(...) is not obligatory in COMP of an embedded clause where subject extraction occurs" (1992b: 48). The optionality of ki in such a position is illustrated in (8). (8)

Ki-moun ou kwè (ki) pral vini? HAITIAN who2sg believe COMP FUT come 'Who do you think will come?' (=(71) in DeGraff 1992b: 48)

To my knowledge, no other author has reported on data similar to those in DeGraff. Dejean (1993: 10) judges as ungrammatical sentences of the type in (8) without the ki. Déprez (1994, 1997) contains an extensive discussion on the distribution of ki proposed by DeGraff (1992b). Further research on the topic will add speakers to the first or the second grammar or possibly to both. In some methodologies (e.g. sociolinguistics), frequency of use of a given form or structure is part of the data, whereas in others (e.g. generative linguistics), it is not. The distribution of ki in Haitian Creole is an example in point. We do not know how many speakers have the respective grammars described above. Scholars report on what they find, and this is how, with time, data on a given language get established. In order to overcome the limits of particular types of databases, researchers may also choose to draw their data from different types of sources. As a general methodological principle, I build databases of different types. My work on Quechua, Montreal non-standard French, and Haitian Creole is based on data elicited in the tradition of field linguistics and generative linguistics, as well as on recorded texts, transcribed and analysed with the assistance of native speakers. For example, in addition to data elicited from some twenty native speakers, my Haitian data also consist of a recorded corpus referred to as the Lefebvre and Fournier (1976) corpus.3 This corpus comprises several tapes of recorded conversations between Haitian immigrants, all workers, who arrived in Montreal between six months to six years prior to the time of the recording. These tapes have all been transcribed (by Robert Fournier), and the transcriptions constitute a source of information against which it is possible to check the grammaticality judgements of the informants in some cases. For example, texts give direct access to surface word order(s) in the subsystems of a creole grammar (e.g. the nominal structure subsystem; the tense, mood and 2 Most of the cited authors work in different cities—some even in different countries. The fact that their informants came up with the same distribution for ki is thus significant. 3 For Quechua, see Lefebvre (1976); for Montreal French, see Doran, Drapeau and Lefebvre (1982).

148

ISSUES IN THE STUDY OF PIDGIN AND CREOLE LANGUAGES

aspect subsystem; etc.). Interpretive data of the type of those in (3) and (4) cannot be checked against texts, for, as was pointed out earlier, texts do not provide this type of data. Texts do not provide researchers with negative evidence like that which is given in (6), nor with data that require syntactic tests (see e.g. (7)). It is worth pointing out that particular structures may not be instantiated in recorded texts. For example, in the Lefebvre and Fournier corpus, there is not a single occurrence of subject extraction involving long-distance Whmovement. Hence, in this case, it is not possible to confront the different grammaticality judgements in (7) and (8) with data produced spontaneously. Nonetheless, whenever possible, checking the grammaticality judgements of speakers against recorded material is a good way of making sure that these two types of data are not in disagreement. Based on my experience, I would say that the two types of data are generally not in disagreement. Apparent problematic cases are discussed in the next section. 5 . 3 . The problem of "inconsistencies" in elicited data The validity of data drawn from elicitation sessions with informants has been challenged on some occasions. For example, Labov (1975) has raised the question of (in)consistencies between informants and within a given informant. I will address these two aspects of the problem in turn on the basis of my own fieldwork experience on Haitian Creole and its source languages. 5.3.1 The problem of inconsistencies between speakers It could be the case that not all speakers of a given language share the same judgements on a given structure. The determiners occurring in the Haitian nominal structure will be used to illustrate this point. It is a well-known fact that count nouns in Haitian Creole may be followed by the nominal determiner, la (or one of its allomorphs); this determiner is both definite and anaphoric.4 The plural marker yo, also occurs postnominally in the context of plural count nouns. According to the informants in Lefebvre (1982: 34), la and yo have the distribution in (9). (9)

4

a. The lexical items la and yo may co-occur within the same nominal structure; in this case, the noun is assigned a definite, anaphoric, plural interpretation. b. The determiner la is not obligatory in the context of yo; when la is not pronounced, the noun is assigned a definite plural interpretation. c. At S-structure, the determiner la must precede the plural marker yo.

The Haitian definite determiner is anaphoric in the sense that it indicates that the information conveyed by the noun phrase is part of the shared knowledge of the participants in the conversation (Fournier 1977; Lefebvre 1982, 1998a; Lefebvre and Massam 1988).

ON DATA

149

These properties are examplified in (10a), (10b), and (10c), respectively. (10)

a. b. c.

tab la yo table DEF PL 'the tables' (in question, that we know of) tab yo table PL 'the tables' tab yo a #table PL DEF #'the tables' (in question, that we know of) 'their table'

HAITIAN HAITIAN HAITIAN

As can be seen in (10c), the surface sequence yo a is possible. All authors agree, however, that, in this case, yo is interpreted as the third person plural possessor (e.g. DeGraff 1992b: 108). All the authors who have discussed the properties of these determiners agree on (9b) and (9c). The property on which authors, and hence Haitian speakers, disagree is the one in (9a), examplified in (10a). The remainder of the discussion will concentrate on this property. When I did the research underlying Lefebvre (1982) on the Haitian determiners, I was not surprised to find that la and yo could co-occur in the grammar of the speakers I worked with, for other authors had already acknowledged this possibility (e.g. d'Ans 1968: 105; Faine 1937: 83; Fournier 1977: 43; Goodman 1964: 45; Sylvain 1936: 55; Valdman 1978: 194-195). Furthermore, there are instantiations of co-occurring la and yo in the Lefebvre and Fournier (1976) corpus. In the literature published after Lefebvre (1982), we find the following reports. Lefebvre and Massam (1988), as well as Ritter (1992: 207-209), also note the possibility of co-occurrence of la and yo. DeGraff (1992b: 107) and Lumsden (1989: 65) report on some Haitian speakers for whom la and yo cannot co-occur.5 Joseph (1988: 201) also reports on some speakers for whom la and yo cannot co-occur, but he specifically mentions that la and yo can co-occur for some speakers; he provides examples of these co occurrences, one of which is reproduced in (11). (11)

Siriyis kraze tab la yo. Siriyis break table DEF PL 'Siriyis broke the tables.'

HAITIAN (=(6.6.b) in Joseph 1988: 201)

Some may consider the above data as inconsistencies between speakers. Others, however, may consider these facts as a case of variation between 5

DeGraff (1992b: 107) and Lumsden (1989: 65) independently propose an analysis in which both la and yo occupy the same position. This analysis cannot hold for speakers who accept the co-occurrence of both morphemes.

150

ISSUES IN THE STUDY OF PIDGIN AND CREOLE LANGUAGES

speakers; from this perspective, there are two slightly different Haitian grammars. These are schematized in (12).6 (12)

a. b.

G1, where la and yo can co-occur G2, where la and yo are mutually exclusive

The assumption that the "variation" view is the correct way of interpreting the data raises the following question: how can we be sure that the facts in (12) represent significant patterns of variation in the language? There are several ways one can confirm the discovery of significant patterns of variation in a language. First, the fact that different authors independently report on the same subsets of data suggests that significant patterns of variation exist. This is the case for the two grammars in (12), found independently by more than one author. A second methodological procedure consists in enlarging one's sample. While doing fieldwork on Haitian and Fongbe, I had recourse to this methodology every time I found variation between a small number of speakers on a given structure. This procedure allowed me to find a limited number (generally two or three) of systematic patterns of variation for a given structure or construction. Examples of such systematic patterns of variation obtained after enlarging my sample of speakers can be found in Lefebvre (1998a: 219-248), for the distribution of the determiner in the clause; pp. 369-370, for the distribution of the determiner in verb doubling constructions, and in Lefebvre (2001a), for the interpretation of deictic terms [see also chapters 3 and 9]. A third way of finding out whether the variation encountered between speakers of a given language is significant or not is to determine if the same patterns exist in a different (but related) language. For example, assuming that Haitian Creole is historically related to Fongbe—(and to other West African languages that were present at the time Haitian Creole was formed), along the lines of Lefebvre (1998a, and the references therein)—, we may ask whether the patterns of variation found in Haitian are also found in its substratum languages. For example, the determiner system of Haitian is quite similar to that of Fongbe (e.g. Brousseau and Lumsden 1992; Lefebvre 1998a: 78-110, and the references therein; Lumsden 1991, and the references therein). Like Haitian la (and its allomorphs), the definite determiner of Fongbe, 5 (and its allomorph), is anaphoric and occurs postnominally. Like Haitian yo, the Fongbe plural marker, lέ, also occurs postnominally. According to the informants in Brousseau and Lumsden (1992) and Lefebvre (1998a), 6 and lέ have the distribution in (13).

6

See also d'Ans (1968: 105), for further discussion of this point.

ON DATA (13)

151

a. The lexical items and lέ may co-occur within the same nominal structure; in this case, the noun is assigned a definite, anaphoric, plural interpretation. b. The determiner is not obligatory in the context of lέ; when is not pronounced, the noun is assigned a definite plural interpretation. c. At S-structure, the determiner

must precede the plural marker lέ.

These properties are illustrated in (14). (14)

a.

b. c.

FONGBE table DEF PL 'the tables' (in question, that we know of) távò lέ table PL 'the tables' ....___ table PL DEF

FONGBE FONGBE

The distribution of the Fongbe morphemes in (13) is extremely similar to that of the corresponding Haitian morphemes in (9). This being the case, do we find, in Fongbe, the two grammars that we find in (12) for Haitian? As can be seen in (14a), the determiner and the plural marker lέ can co-occur. This corresponds to G1 in Haitian (see (12a)). Now, are there Fongbe speakers for whom and lέ cannot co-occur; that is, Fongbe speakers who have a grammar corresponding to G 2 in Haitian (see (12b))? For all twenty-five Fongbe speakers that I worked with, and lέ may co-occur. In the literature, however, there is one author (also a native speaker of the language) who points out that, for some speakers, the determiner cannot occur in the environment of lέ (Agbidinoukoun 1991: 149).7 This shows that, as is the case in Haitian, there are two slightly different grammars of Fongbe with respect to the possibility of co-occurrence of the definite determiner and the plural marker. These two grammars are represented in (15). (15)

a. b.

G1, where and lέ can co-occur G2, where and lέ are mutually exclusive

The fact that different languages with a similar determiner system present the same type of variation in this system suggests that the variation found in each of these languages taken separately may be significant, regardless of the size of the 7

This suggests that, if we were to have a larger sample of speakers, we might find speakers for whom and lέ cannot co-occur.

152

ISSUES IN THE STUDY OF PIDGIN AND CREOLE LANGUAGES

sample of speakers who manifest what would appear to be a less frequent pattern (in this case, G 2 in both (12b) and (15b)).8 There are thus several strategies that one can use to find out whether the variation encountered between the speakers of a given language is significant. In my experience, what may appear at first glance to be inconsistencies between speakers generally turns out to constitute consistent patterns of variation in the language. 5.3.2 The problem of inconsistencies within a given speaker When a speaker is questioned on the properties of a given structure more than once, (s)he may produce the same judgements the second or the third time. In this case, the speaker is considered to have consistent judgements. It may happen, however, that a speaker will not have the same judgements the second or the third time. This situation has been used by some authors to illustrate the weaknesses of databases that rely on grammaticality judgements. For example, Labov writes: "Lack of consistency on repeated testing has been among the most serious weaknesses of the claims for idiosyncratic dialects" (1975: 27). As part of the methodology I have established for my fieldwork, I always test the grammaticality judgements of each speaker at least twice (sometimes more) for a given construction. Over and over again, I have found that speakers were quite consistent in their grammaticality judgements; that is, they generally had the same evaluation of the data on the repeated testing. There are a few cases, however, where I kept getting slightly different interpretations for the same sentences from session to session. In such cases, there are two options. The first one is to conclude that the speaker is inconsistent, and sometimes, this is the right conclusion. The second one is to try to find out why there is this variation in individual speakers' judgements and to use this variation as a trigger for a new way of looking at the data. For example, while I was working on the Haitian clausal determiner (Lefebvre 1998a: 219-248, and the references therein), I kept registering slightly different interpretations of the data from the same speaker from one session to the next. My first evaluation of this situation was that the speaker was inconsistent in his judgements. I then started to work with other speakers. To my surprise, I found the same so-called inconsistencies with these other speakers as well. At this point, I decided to look at the data from a different point of view and to consider 8 Some scholars might want to argue that significant patterns of variation require detailed sociolinguistic studies. While I would agree with such a statement at some level of analysis, the discussion in this section is set within the framework of another paradigm. I would say that the slightly different grammars in (12) and (15) reveal significant patterns of variation between speakers. These grammatical patterns may be considered as a "variable" (in the sense of Labov 1972a) for community studies.

ON DATA

153

that there was some order in what first looked like inconsistencies. This turned out to be a successful methodology, for it led me to propose that the Haitian clausal determiner can occupy more than one position in the syntactic tree. Each position that can host this lexical item determines its scope; this explains the slightly different interpretations of the sentences containing the clausal determiner. Had I abandoned the topic at an early stage of my investigation, I would have missed one very important property of the clausal determiner, that of being able to occur in more than one syntactic position in the clause structure.9 5 . 4 . Native speakers as informants It is worth noting that being a native speaker of a given language does not necessarily confer one with clear judgements on one's grammar. Some people are naturally good informants, while other people are not. Like in other areas of life, being a good informant—that is, having clear judgements on one's grammar—is a matter of talent or skill, regardless of the language under consideration. Researchers have to be able to identify speakers with such a skill, whether they work on a creole language or not. In the case of creole communities, the situation is extremely complex for creole native speakers may speak different varieties that are quite far apart from one another. For example, the Haitian varieties referred to as basilectal, mesolectal, and acrolectal do not always present the same data (Valdman 1978: 292-295). Speakers of different varieties may have different judgements on subsets of data. On the basis of findings on varieties of English, Labov (1972a) has convincingly shown that speakers of a given variety do not have direct access to the grammar of speakers of the other varieties. Differences between grammaticality judgements among Haitian speakers may be attributable to this dimension of variation. It is possible that the variation found in the distribution of ki, discussed earlier (see (7) and (8)), is an example of such a type of variation. (For another interpretation of this variation, see Déprez 1994.) Furthermore, there is variation between speakers of the same social group, as community studies have shown us over and over again (e.g. Labov 1972a, 1972b). Community studies on Haitian Creole will most certainly reveal examples of this type of variation. Finally, there is regional variation, a topic that has received some attention in the literature on Haitian Creole (e.g. Fattier 1998; Orjala 1970; Tinelli 1970). A classic example of this type of variation in Haitian Creole involves the manifestation of case on the possessor phrase in the nominal structure. While the northern part of Haiti uses a prenommai possessive particle, the central and southern parts of the country have a phonologically null case 9

See also Cowart (1998), for the elaboration of a methodology that can tell us whether the variation within a given speaker is significant or not.

154

ISSUES IN THE STUDY OF PIDGIN AND CREOLE LANGUAGES

marker. This is shown in (16), from Lumsden and Lefebvre (1994: 109), based on e.g. Goodman (1964), Lumsden (1991), and Orjala (1970: 36). (16)

a. b.

liv a Jan book PART John 'John's book' liv Jan 0 book John GEN 'John's book'

NORTHERN HAITIAN CENTRAL AND SOUTHERN HAITIAN (=(1) in Lumsden and Lefebvre 1994)

Given the data in (16), it is to be expected that speakers of Haitian will have different grammaticality judgements with respect to case marking of the possessor phrase. Given all these dimensions of potential variation in the data, it is not surprising that data provided by some speakers may, in some cases, contradict those of other speakers of the language. Researchers have to take variation into account when they are collecting their data. Moreover, their choice of basilectal, mesolectal, or acrolectal speakers as informants is determined by their research questions, as mentionned in the first section of this column. 5 . 5 . Can data on a given language ever be complete? Throughout this column, we have seen that data collection is not a neutral activity and that the type of data researchers gather is not independent from their research paradigms and questions. Establishing data on a given language is thus a collective entreprise. This is how, with time, data on a given language and patterns of variation in a language become established. Can data on a given language ever be complete? In light of the above discussion, the answer to this question cannot be positive. The task of establishing the data for any language is endless.

CHAPTER 6

MULTIFUNCTIONALITY AND THE CONCEPT OF LEXICAL ENTRY* A lexical item that fulfils more than one grammatical function is multifunctional. For example, the lexical item táki in Saramaccan and in Sranan has the function of a verb meaning 'to say' and that of a THAT-type complementiser. In this chapter, I will be concerned exclusively with such word base multifunctional items.1 Multifunctionality is a widespread phenomenon. As will be shown in the second section of this column, it is a property of all human lexicons. The range of multifunctional lexical items in particular lexicons may vary but it is not negligible. For example, in Valdman's et al. (1981) dictionary, more than three hundred Haitian Creole lexical items are identified as having more than one function. Multifunctional items range across all syntactic categories, as will be shown in the second section of this column. The phenomenon calls into question the concept of lexical entry. Because it is at the intersection of lexicon, semantics and syntax, it raises a number of theoretical questions such as the following. How can multifunctionality be characterised? What is an optimal account of the phenomenon? Multifunctionality has been of great interest to creolists, as one can tell from the number of documented cases in the literature over the last twenty years.2 Furthermore, multifunctionality and the process of grammaticalisation are The content of this chapter was originally published as my third column for the Journal of Pidgin and Creole Languages (Lefebvre 200 1b). Research that led to the content of this column was financed by SSHRCC, FCAR, and FIR (UQAM). I would like to thank Anne-Marie Brousseau, Catherine Léger, Ingo Plag, Jean-François Prunet and Lisa Travis for their comments and questions on an earlier draft of this column. I also want to thank Renée Lambert for her assistance in documenting some of the points raised in this column and Andrée Bélanger for formatting it. 1 Consequently, root base cases, as well as cases which require some morphology as in Mühlhäusler (1985), are not considered here. 2 Cases in point are: On Belisean creole: Escure (1993); On Gullah: Mufwene (1993); On Ndjuka: Huttar and Koanting (1993); On Negerhollands: Hinskens and van Rossem (1995); On Nigerian Pidgin: Mann (1993); On Saramaccan: Byrne (1987); Veenstra (1996a, 1996b);

156

ISSUES IN THE STUDY OF PIDGIN AND CREOLE LANGUAGES

not unrelated, for grammaticalisation may be viewed as a process creating multifunctional lexical items. The topic of grammaticalisation has received much attention in recent literature on creoles3 and on other languages as well.4 Moreover, in recent publications there has been some discussion on the source of multifunctional lexical items in creole languages. In this respect, the contribution of the creole source languages has been investigated, a topic which will be taken up in section 6.8. For all these reasons, the topic of multifunctionality (and related phenomena) is timely. There are several perspectives from which one can address this topic and, to some extent, some aspects of the problem depend on one's theoretical perspective. Nonetheless, I believe that the basic questions raised by the problem remain the same regardless of the point of view of the researcher. In this column, I address the topic of multifunctionality as it relates to the concept of lexical entry. I will assume a concept of lexical entry where lexical entries are specified for semantic and syntactic features. I will also assume that lexical entries may be underspecified for both semantic and syntactic features. I explore the view that underspecification is the property that accounts for the multifunctional character of lexical items. Underspecification may be implemented in various ways. In this column, I address the matter in terms of semantic and syntactic features. Other approaches will be mentioned in the text. The content of the column is organised in the following way. First, I consider the tests that allow for the identification of multifunctional lexical items showing that each case needs to be argued for. Then, I consider the range of multifunctional lexical items across syntactic categories showing that multifunctional items manifest themselves across all types of syntactic categories. The next three sections are dedicated to the account of multifunctionality in terms of semantic and syntactic underspecification, motivated by the monosemy principle. The question of the relationship between multifunctionality and grammaticalisation is then taken up in light of this approach. I consider the problem of apparent cases of grammaticalisation and that of the transfer of substratum multifunctional lexical items into a creole. Alternative views and topics for future research are identified throughout.

On Seychellois: Gilman (1993); On Sranan: Arends (1986); Bruyn (1995, 1996); Plag (1993); Voorhoeve (1981); Winford (1993); On Tok Pisin: Mühlhäusler (1985); Romaine (1992); Sankoff (1991, 1993); etc. 3 See e.g. Baker and Syea (1991); Baker and Syea (eds) (1996), and the papers therein; most of the papers cited in note 2; Plag (1995); Sankoff (1980); Sankoff and Brown (1980); Sankoff and Laberge (1980); etc. 4 See e.g. Pagliuca (ed.) (1994); Traugott and Heine (eds) (1991), vol.1 and vol. 2.

MULTIFUNCTIONALITY AND THE CONCEPT OF LEXICAL ENTRY

157

6.1.

Establishing the various functions of multifunctional lexical items Syntactic tests distinguish nouns from adjectives, verbs from prepositions, etc. The syntactic category of a given lexical item is established on the basis of syntactic tests. Given this fact, a multifunctional lexical item is one that can be shown to have the properties of more than one syntactic category; that is, it sometimes manifests the properties of one category and sometimes those of another one. The multifunctional character of individual lexical items needs to be established on the basis of syntactic tests. Unless otherwise specified, all the examples in this section are from Haitian Creole. For example, the lexical item bezwen, 'need' or 'to need', sometimes behaves as a noun; that is, it can manifest the syntactic properties of nouns. In this case, like other nouns, it can be preceded by the modifier anpil 'a lot'. This same lexical item can also behave like a verb; that is, it can manifest the syntactic properties of verbs. In this case, like other verbs, it can take a subject, it can be preceded by a TMA marker (as in Mwe te bezwen lajan T needed money'), and it can participate in verb doubling constructions, as is illustrated in (l). 5 (1)

Se bezwen it.is need fè' m cause me 'It is the fact

m' bezwen lajan an, ki I need money DEF RES dwe travay. have.to work that I need money that causes me to work.'

HAITIAN

The lexical item ansyen 'old' can behave as a count noun, which can be marked for plural (e.g. ansyen yo 'the old people'). It can also behave like an adjective. In this case, like other lexical items of this syntactic category, it can modify a noun (as in ansyen kay la 'the old house'; see also Joseph 1999). The lexical item kay can function as a noun meaning 'house'; it can also function as a preposition, as shown in (2). (2)

Pa vin fè bri kay moun yo. HAITIAN not come make noise at.the.home.of person PL 'Do not come to make noise at the home of people.' (from Valdman et al, 1981)

Like other prepositions, kay in (2) obligatorily selects a complement, in this case an NP. As is the case with the other prepositions in the language, kay and its complement form a constituent that can be questioned, as shown in (3 a). Like other prepositions, kay cannot be stranded, as is shown by the ungrammaticality of (3b). 5

Unless otherwise specified, the examples cited in the text are from my own data.

158

ISSUES IN THE STUDY OF PIDGIN AND CREOLE LANGUAGES

(3) a. Ki kote i ou dwe ate ti ? Kay Pyè which place you have.to go at.the.home.of Peter 'Where do you have to go? At the home of Peter.' b . *ki mourn i ou dwe ate kay ti which person you have.to go at.the.home.of

HAITIAN

HAITIAN

While the object of prepositions cannot be extracted out of the prepositional phrase, as shown in (3b), the object of verbs can be extracted out of the verbal phrase. This test is used by Déchaine (1988) to distinguish the two functions of baylba6; that is, to distinguish its function as a verb meaning 'to give' from its funciton as a benefactive preposition meaning 'for'. As is shown in (4a), the object of the verb bay 'to give' may be extracted (in this case questioned). However, the object of the preposition ba 'for' cannot, as is shown by the ungrammaticality of (4b). (4) a. Ki mourii Jan pran dlo bay ti ? HAITIAN which person John take water give 'To whom did John give water?' (=(33) in Déchaine 1988: 43) b . *ki mourii li te pote liv sa a ba ti HAITIAN which person (s)he ANT bring book DEM DEF for [Lit.: 'Whom did (s)he bring this book for?'] (=(34a) in Déchaine 1988: 44) Another test that distinguishes the two categorial uses of baylba is the following: in its function as a verb, baylba and its object cannot be fronted, as shown by the ungrammaticality of the sentence in (5a), but as a preposition, baylba and its object can be fronted, as shown in (5b). HAITIAN *bay ki mourn i Jan pran dlo ti give which person John take water (=(35) in Déchaine 1988: 44) b . Ba ki mouni li te pote liv sa a ti HAITIAN for which person (s)he ANT bring book DEM DEF 'For whom did (s)he bring this book?' (=(35) in Déchaine 1988: 44)

(5) a.

6

Several authors (see e.g. Hall 1953; Valdman 1996; Valdman et al 1981) list the forms bay and ba as allomorphs. Sylvain (1936: 104) reports that the full form is used when the lexical item is not followed by its object, as in (i), whereas the short form is used when the lexical items is followed by its object, as in (ii). (i) kasav mwen bay HAITIAN casava I give 'the casava that I gave' (from Sylvain 1936: 104) (ii) Ba pitit yo manje. HAITIAN give child PL eat 'Give the children something to eat.' (from Sylvain 1936: 104) For a preliminary discussion of this alternation, see also Déchaine (1988: 46, 48).

MULTIFUNCTIONALITY AND THE CONCEPT OF LEXICAL ENTRY

159

The two tests discussed above illustrate how syntactic categories are established for major lexical items. Syntactic tests can also distinguish lexical categories from functional categories. For example, the Haitian lexical item pase has the meaning 'to pass' in (6a) and 'to surpass' in (6b). (6)

a. b.

Jan pase kite Mari. HAITIAN John pass leave Mary 'John surpassed Mary.' Moun sa a (plis) byen pase nou. HAITIAN person DEM DEF more happy surpass us 'This person is happier than us.' (=(50) in Déchaine 1988: 48)

According to Valdman et al. (1981), pase in (6b) is a conjunction rather than a verb. Likewise, Déchaine (1988: 50) identifies it as a comparative marker. Contra Jensen et al. (1978), she argues that the lexical item pase occurring in comparative constructions, as in (6b), is not a (serial) verb. A test demonstrating that pase is not the second verb of a serial verb construction is the following: while it is possible to extract the object of the second verb of a series, it is not possible to extract the object of pase in comparative constructions, as is shown by the ungrammaticality of (7). (7)

*se nou moun sa a (plis) byen pase it.is us person DEM DEF more happy surpass

HAITIAN

Another similar example consists of a lexical item that can serve both as a verb and as a THAT-type complementiser. Saramaccan and Sranan, but not Haitian Creole, have such a lexical item. The sentences in (8a) and (8b) provide examples of the two forms of the Saramaccan verb meaning 'to say': taki/táa; the sentences in (8c) and (8d) show that the same forms may be used as a complementiser (for further discussion of these forms and of co-occurrence restrictions between them, see e.g. Byrne 1987; Veenstra 1996a). (8)

a. b. c.

Hén a táki: mi nángó. SARAMACCAN and.then 3sg say lsg ASP=go 'And then he said: I am leaving.' (=(2a) in Veenstra 1996a: 155) Hén a táa mi nángó. SARAMACCAN and.then 3sg say lsg ASP=go 'And then he said: I am leaving.' (=(la) in Veenstra 1996a: 155) A táki táa a náki hén. SARAMACCAN 3sg say say 3sg hit 3sg 'He said that he had hit her.' (=(2b) in Veenstra 1996a: 155)

160

ISSUES IN THE STUDY OF PIDGIN AND CREOLE LANGUAGES d.

Mi sábi táa á búnu. SARAMACCAN lsg know say 3sg=NEG good 'I know that he is not good.' (=(3b) in Veenstra 1996a: 155)

The sentences in (9a) and (9b) illustrate similar facts for Sranan (see also Bruyn 1996; Mufwene 1996a; Plag 1995; Sebba 1987; Winford 1993). (9) a. Da' Anansi taki, eng go proben efi eng kan saki fa a nem. 'Then Anansi said, he would try if he could learn his name.' (=(8a) in Plag b. Mi sabi taki Marlon no fufuru na fowru. I know SAY Marlon NEG steal the chicken T know that Marlon didn't steal the chicken.' (=(la) in Plag

SRANAN 1993: 40) SRANAN 1993: 36)

A major test, from the aforementioned references, that distinguishes between the two functions of taki/táa is the following. As a verb, taki/táa may participate in the verb doubling phenomena, but as a complementiser it cannot. This is shown by the contrast in grammaticality between the (a) and the (b) sentences in (10). (10) a.

b.

Verb SARAMACCAN Táa a táa: m'e nángó. say 3sg say lsg=NEG ASP=go 'He really said: I am not leaving.' (=(6b) in Veenstra 1996a: 156) Complementiser *táa a táki táa á búnu SARAMACCAN say 3sg say say 3sg=NEG good (=(8a) in Veenstra 1996a: 157)

A similar test is provided for Sranan. Plag (1993: 42) mentions that verbs may undergo predicate clefting. He shows that as a complementiser taki cannot be clefted, as illustrated in (11). (11) a. b.

Anansi taigi a uma taki, eng sref sa bori. SRANAN 'He said to his wife that he himself would cook.' (=(13a)inPlagl993:42) *Na taki Anansi taigi a uma taki eng sref sa bori SRANAN (=(13b)inPlagl993:42)

The examples provided in this section illustrate how categorial properties are established for particular lexical entries. Further tests can be found in the literature. The important point here is that one must have syntactic tests to show that words can act syntactically like one category in some cases and like another one in other cases. Although all researchers might not agree on the account of this type of phenomena, in most cases, any account has to allow for some sort of multifunctionality.

MULTIFUNCTIONALITY AND THE CONCEPT OF LEXICAL ENTRY

161

6.2.

The range of multifunctional lexical items across syntactic categories One may wonder about the range of multifunctional lexical items across syntactic categories. Is multifunctionality restricted to a few syntactic categories or is it a phenomenon that can take place across the whole range of syntactic categories? On the basis of Haitian Creole, I show that multifunctional lexical items cover a whole range of syntactic categories. The words in (12) can all be used as nouns. Some can also be used as verbs, as in (12a, a'); others as adjectives, as in (12b); and yet others as prepositions, as in (12c, c', c"). (12) a. a', b. c. c'. c".

apèsi

noun verb bezwen noun verb ansyen noun adjective granmèsi noun preposition kay noun preposition kote noun preposition

'glimpse' HAITIAN 'to notice' 'need' HAITIAN 'to need' 'old person' HAITIAN 'old' 'thanks' HAITIAN 'thanks to' 'house' HAITIAN 'at the house of' 'place/side' HAITIAN 'besides/near' (from Valdman et al. 1981)

The examples in (13) consist of lexical items that can all be used as verbs. Some can also be used as nouns, as in (13a, a'); others as adjectives, as in (13b); and yet others as prepositions, as in (13c). (13) a.

koze

a'.

sdti

b.

gate

c.

pran

verb 'to chat' HAITIAN noun 'chat' verb 'to go out (from), to come from' HAITIAN noun 'stroll, drive (outing)' verb 'to rot' HAITIAN adjective 'rotten (spoiled)' verb 'to take' HAITIAN preposition 'as of' (from Valdman et al 1981)

The lexical entries in (14) can all be used as adjectives. As is shown in (14a) and (14b), respectively, these lexical items may also be used as adverbs and stative verbs. (14) a.

enjis

adjective adverb

'unfair' 'unfairly'

HAITIAN

162

ISSUES IN THE STUDY OF PIDGIN AND CREOLE LANGUAGES b.

wo

adjective adverb verb

'tall' 'high' 'to be tall'

HAITIAN

All the lexical entries in (15) can be used as prepositions. The examples in (15a), illustrate that the prepositions antè may also serve as a noun, an adjective or an adverb. The examples in (15b) and (15c) show a similar pattern for the prepositions devan and anwo. In (15d), we see that the preposition andedan may also be used as a noun, while in (15e), the preposition anba may also be used as an adverb. (15) a.

ante

b.

devan

c.

anwo

d.

andedan

e.

anba

preposition noun adjective adverb preposition noun adverb preposition noun adverb preposition noun adverb preposition adverb

'on top of, above' HAITIAN 'top, upper part' 'flighty' 'high, up high' 'before, in front of' HAITIAN 'front, lead (position)' 'ahead, in front' 'on, on top of HAITIAN 'top' 'up there, upstairs' 'inside' HAITIAN 'inside' 'inside' 'under, below' HAITIAN 'below, underneath' (from Valdman et al 1981)

The data in (12)—(15) thus show that lexical items of one major syntactic category may also be used with the properties of at least one other major syntactic category. The examples in (16)—(18) illustrate major category lexical items that can also be used as functional items. For example, the data in (16) consist of prepositions that may also be used as conjunctions. (16) a. b.

depi

preposition conjunction tankou preposition conjunction

'since' HAITIAN 'since/as soon as' 'like, just like' HAITIAN 'just like, like' (from Valdman et al 1981)

Likewise, Haitian pou may function as a preposition meaning 'for'. As such, it may select a nominal or a clausal complement, as is shown in (17a) and (17b). This lexical item may also serve as an irrealis mood marker, illustrated in (17c), or a complementiser, selected by verbs of the WANT-class or by verbs of the BE

MULTIFUNCTIONALITY AND THE CONCEPT OF LEXICAL ENTRY

163

GOOD-class, as is illustrated in (17d) and (17e), respectively (for tests establishing the various functions of Haitian pou see Koopman and Lefebvre 1981, 1982; Lefebvre 1993a, 1998a; Sterlin 1988, 1989). (17) a.

b.

c.

d.

e.

Preposition selecting a nominal complement Pote sa pou mwen. HAITIAN bring DEM for me 'Bring this for me.' (=(4) in Koopman and Lefebvre 1981) Preposition selecting a clausal complement M' te bezwejt èskont lè sa-a HAITIAN I ANT need money time DEM-DEF pou m' te reparti. for I ANT go 'I needed money then in order to go back.' (=(6) in Koopman and Lefebvre 1981) Irrealis mood marker Tout solda pou vini laplas. HAITIAN all soldier MO come square 'All soldiers have to come to the square.' (from Sylvain 1936: 90) Complementiser selected by verbs of the WANT-class Yo te vle [pou m' te antre HAITIAN they ANT want COMP I ANT join nan troup Jakmèl]. in troop Jacmel 'They wanted me to join Jacmel's troops.' [Lit.: 'They wanted that I joined Jacmel's troops.'] (=(10) in Koopman and Lefebvre 1982) Complementiser selected by verbs of the BE GOOD-class Li bon [pou m t a pati]. HAITIAN it good COMP I ANT IND.FUT leave 'It is good that I leave.' (=(65) in Lefebvre 1993a)

Lexical items that may be used as a verb and as a complementiser (see (8) and (9)) or as a verb and as a comparative marker (see (6)) are of the same type. Furthermore, Gilles (1988) shows that the Haitian prepositions nan 'in' and sou 'on' may also function as case markers in some contexts (see also Joseph 1994, on the ambiguous categorial status of nan). From a descriptive point of view, the Haitian morpheme yo in (18) constitutes a similar example for it is the form of both the third-person plural personal pronoun, a major category lexical entry (see (18a)), and the plural marker, a functional category, occurring in nominal structures (see (18b)). (18) a.

Yo vini. they come 'They came.'

HAITIAN

164

ISSUES IN THE STUDY OF PIDGIN AND CREOLE LANGUAGES b.

pitit yo child PL 'the children'

HAITIAN

The examples from (16)—(18) all illustrate cases of major category lexical items that can also be used as functional items. Finally, there are functional items that may have more than one grammatical function. Haitian ap is a case in point. Indeed, this morpheme has been shown to serve both as an imperfective marker with dynamic verbs and a definite-future marker with verbs of all aspectual classes (see e.g. Lefebvre 1996a, 1998a; Lumsden 1995a; Spears 1990; Valdman 1970). This is illustrated in (19). (19) a.

b.

Imperfective aspect marker Man ap manje krab la. Mary IMP eat crab DEF 'Mary is eating the crab.' Definite future marker M ap vini. I DEF.FUT come T will (definitely) come.'

HAITIAN (=(18) in Lefebvre 1996a) HAITIAN (=(7) in Lefebvre 1996a)

In Lefebvre (1998b), it is argued that the definite determiner is also a multifunctional item in Haitian. This lexical item will be further discussed below, in the section on categorial (under)specification. As can be seen from the examples above, multifunctionality is not restricted to a subclass of syntactic categories. The phenomenon may involve members of all major syntactic categories, and it may involve lexical as well as functional lexical items. The examples provided in this section show that, from a categorial point of view, multifunctional lexical items divide into three groups: one in which a given lexical item may belong to more than one major category (see (12)—(15)), one in which a given lexical item may serve both as a major category item and a functional category item (see (16)—(18)), and another one in which a given lexical item may belong to more than one functional category (see (19)).7 Based on the examples given so far, it seems that multifunctional lexical items do not constitute a marginal phenomenon, for they appear across all type of syntactic categories. It could be argued, however, that all the examples provided so far come from creole languages (Haitian, Sranan and Saramaccan) and that creole languages manifest a wider range of multifunctional items than non-creole languages. I doubt that this would be the case, however, as the following examples, drawn from languages not known to be creoles, show. 7

More cases of this third type need to be documented.

MULTIFUNCTIONALITY AND THE CONCEPT OF LEXICAL ENTRY

165

For example, nouns that are used as verbs are found in English (see e.g. Clark and Clark 1979) and in Tukang Besi (Donohue 1999a), to name only these two. Nouns that are used as adjectives are numerous in French and in English. They are also found in Fongbe (Lefebvre and Brousseau 2002). Nouns that are used as prepositions are found in French and English, and nouns that are used as postposition are found in Fongbe (Lefebvre and Brousseau 2002), in Quechua (Lefebvre and Muysken 1988), etc. Major category lexical items that may also be used as functional items are found in many languages. The verb meaning 'to say' is also used as a complementiser in several West African languages (e.g. Lord 1976). In Fongbe, the preposition meaning 'for' is also used as a complementiser (e.g. Lefebvre 1998a). In English, the deictic term that is also used as a complementiser. The French preposition à also serves as a case marker (see e.g. Kayne 1984). Finally, functional lexical items that may serve more than one grammatical function are also found in non-creole languages. For example, in Fongbe the same lexical item serves both as a mood marker and as a complementiser (e.g. Lefebvre 1998a). In Igbo, the same marker is used for progressive and future (e.g. Déchaine 1991).8 It thus appears that there are multifunctional items in all the languages of the world. Thus, multifunctionality is not a marginal phenomenon. Rather, it appears to be a characteristic of human lexicons. How is multifunctionality compatible with the concept of lexical entry? 6.3.

Multifunctionality, the monosemy principle and semantic under specification In the discussion that follows, I will assume a concept of lexical entry where lexical items are minimally defined for semantic and categorial features. As we saw above, however, a single form may belong to more than one syntactic category and may have slightly different, though related, meanings depending on which of its functions is being considered. Is multifunctionality compatible with the concept of lexical entry? The answer to this question depends, of course, on the concept of lexical entry that is adopted. A first concept of lexical entry can be defined as follows: lexical entries are precisely defined for semantic features and fully specified for syntactic features. In this view, lexical items having the same form are registered in the lexicon as separate lexical entries. Following this concept, there would be two lexical entries for Haitian ba(y). One would consist of ba(y), categorially identified as a verb and semantically specified as meaning 'to give'; the other one 8 For a discussion of several types of multifunctional categories based on non-creole languages, see e.g. the papers in the Revue québécoise de linguistique 27 (2), 1999, a special issue on Syntactic categories, and the references therein.

166

ISSUES IN THE STUDY OF PIDGIN AND CREOLE LANGUAGES

would consist of ba(y), categorially specified as a preposition and semantically defined as meaning 'for'. This position is the one adopted in e.g. Veenstra (1996a, 1996b), who posits two lexical entries for the Saramaccan form mbéi: a verb meaning 'to make' and a complementiser meaning 'so that', three lexical entries for dá: a verb meaning 'to give', a "dative" preposition and a prepositional complementiser-like meaning 'so that', and two lexical entries for táa a verb meaning 'to say' and a THAT-type complementiser. A second concept of lexical entry can be defined as follows: a given form is registered in the lexicon as one lexical entry; the semantic and syntactic features of this lexical entry are defined in such a way as to encompass the multiple, though related, meanings and the different syntactic functions of this lexical entry. This concept of lexical entry is based on the monosemy principle. In recent literature (see e.g. Bouchard 1995; Cowper 1989, 1995; Ghomeshi and Massam 1994; Johns 1992; Nida 1948; Ruhl 1989; etc.), it has been proposed that monosemy is to be preferred over polysemy wherever possible. This principle is known as the one form/one meaning principle in Johns (1992: 84): "Where morphemes are identical or similar in phonological properties, in the unmarked case, they are identical or similar in all lexical properties". As is pointed out in Lefebvre (1999a), this general principle does not apply to forms that came to have the same phonological representation as the result of diachronic phonetic changes. For example, in modern French, the words verre 'glass', vers 'towards', and vert 'green' have the same phonological representation due to historical developments in the pronunciation of French. Likewise, in Lefebvre (1999a), it is assumed that this principle does not apply to Haitian forms derived from phonetically similar, but semantically and historically unrelated, French forms that ended up having identical phonological representations. The Haitian form siye in (20) is a case in point. (20)

/siye/ /siye/ /siye/

'to saw'
'to saw' HAITIAN 'to wipe' 'to associate with' (from Valdman et al. 1981) The French and the Haitian examples above constitute cases of homonymy or homophony and need not concern us here. According to the monosemy principle, one should avoid proposing several lexical entries for lexical items having the same phonological form if the meanings corresponding to these lexical items are related. Thus, ba(y) in (4) and (5) would be registered as one lexical entry, while siye in (20) would be registered as multiple lexical entries.

MULTIFUNCTIONALITY AND THE CONCEPT OF LEXICAL ENTRY

167

In current literature, the semantics of a lexical entry is given in the form of a lexical conceptual structure (LCS). According to the definition in Hale and Keyser (1987: 5), an LCS "is a representation of the concept named by the verb, and the 'participants' in the action, represented by 'variables', corresponding to the TH(ematic)-roles assigned to grammatical functions". In this respect, a more precise statement of the monosemy principle is formulated by Cowper (1995: 3) in terms of the strong monosemy principle: "The conceptual structure of a lexical entry may contain no disjunctions and no optional elements. If the conceptual structure of two uses of a lexical item cannot be unified through underspecification, then they must be treated as distinct lexical entries".9 In other words, related meanings may be expressed as part of the same lexical entry by means of the semantic underspecification of a LCS. Ruhl (1989: 234) suggests the following methodology: "Assume that any meaning that is not present in all contexts of a word is not part of the word's inherent meaning; if this fails, assume distinct meanings are figuratively related". Thus, the various meanings of a given lexical entry must have a semantic core which can be represented by a single lexical conceptual structure. For example, for a verb that has several related meanings, the LCS of the verb will be underspecified so as to subsume the verb's more specific meanings. A relevant example is provided by the Haitian verb kenbe, which means 'to hold' and 'to hold on', 'to keep in containment', 'to capture' or 'to bring into containment' and 'to contain' (see Lumsden 1995a; Valdman et al. 1981). The LCS of this verb should thus be abstract enough so as to subsume all of its related meanings. In contrast, for verbs whose meanings are more specific, the LCS s should also be more specific so as to reduce the range of contexts in which these verbs can appear. For example, the LCS of to cut in English allows this verb to be used in various contexts. Thus, to cut can take fruit, meat, or hair as objects. Other languages, however, have several verbs encoding the meaning 'to cut'. In Fongbe, for example, there are different verbs of cutting whose meanings depend on the object being cut. Hence 'to cut hair' requires the verb kpà, 'to cut a banana' requires the verb mà, 'to cut meat' requires the verb kàn, etc. (see Brousseau 1993, 1998; Lefebvre 1989, 1999a). The LCSs of these Fongbe verbs will thus have to be formulated in a way specific enough to limit their use to the appropriate object (and thus to the appropriate manner of cutting).10 The exact formulation of LCSs for both under- and fully- specified LCSs still needs to be worked out. 9

See also Bouchard (1995) and Ruhl (1989) for an extensive discussion of this topic. On the assumption that the implementation of the monosemy principle described in the text is correct, the observation in Romaine (1992: 151) according to which multifunctionality "violates the principle of one form equals one meaning" needs to be revised. 10

168

ISSUES IN THE STUDY OF PIDGIN AND CREOLE LANGUAGES

The examples discussed so far in this section are cases where a given form may have several (related) meanings but only one syntactic function, as the above examples all involve verbs. The next section discusses cases where, in addition to having related meanings, a given lexical item may belong to more than one syntactic category. In addition to being semantically underspecified, a lexical item of this kind will need to be categorially underspecified. 6.4. Categorial under specification 6.4.1 Theoretical assumptions In the discussion that follows, I will assume a theory of major syntactic categories where categories are defined by the major features [aN, ßv], where a and ß are variables that can take the value + or - (see e.g. Chomsky 1972, 1974; Jackendoff 1976, 1977). The combinations of these features yields the four major categories listed in (21). (21)

[+N,-V] [+N, +v] [-N, +v] [-N, -V]

= = = =

noun adjective verb preposition

I will further assume that functional lexical items are defined by minor features such as [±T(ense)], [±Wh], [±DEF(inite)], etc. In this view, a lexical item X is specified for a subset of the syntactic features in (22). (22)

X=[±N, ±V], [±T], [±Wh], [±DEF], etc.

Given this framework, multifunctional lexical entries will be underspecified for categorial features. Categorial underspecification allows a given lexical item to be used as a member of more than one syntactic category. The underspecification of major and minor categories will be discussed in turn.11 6.4.2 The underspecification of major category lexical items A major category lexical item may be underspecified for the nominal feature; in this case, it is [aN], where a can take the value + or -. Thus, a lexical item bearing the features [aN, -V] will be a noun when a is assigned the value +, and it will be a preposition when a is assigned the value -. When either [aN] or [ßv] (but not both) is underspecified, the combinations listed in (23) obtain.

11 To my knowledge, a holistic theory of syntactic features is not available as yet. Consequently, a theory of underspecification of features is not available either. The content of this section should thus be understood as aiming at some formal representation of categorial underspecification.

MULTIFUNCTIONALITY AND THE CONCEPT OF LEXICAL ENTRY

169

Each combination corresponds to a case of multifunctionality reported on earlier in this column. (23) a. b. c. d.

[αN, -V]: [αN, +v]: [+N, ßv]: [-N, ßv]:

[+N, -V] (N) or [-N, -V] (P) [-N, +V] (V) or [+N, +v] (A) [+N, -V] (N) or [+N, +V] (A) [-N,+V] (V) or [-N,-V] (P)

(see (2)) (see (13b)) (see (12b)) (see(4) and (5))

The underspecification for feature value in (23) accounts in a straightforward way for the multifunctional character of the major category lexical items identified above. The above discussion raises the question of whether all major categories can give rise to multifunctionality. This amounts to asking whether underspecification is a multidirectional phenomenon.12 In (23), we saw that multifunctionality may involve the pairs of syntactic categories in (24). (24) a. b.

N or P N or A

c. d.

V or A V or P

According to (24), all major category lexical items can be multifunctional. However, in (24), nouns and verbs are basic, in the sense that, in addition to being nouns and verbs, respectively, they may also be used as prepositions or adjectives. Can adjectives and prepositions be basic in the same way? That is, can adjectives also be used as verbs or prepositions; can prepositions also be used as nouns, verbs or adjectives? This appears to be impossible as is schematically presented in (25). (25) a. b. c.

*A or V *A or P *P or N

d. e.

*P or V 1 3 *P or A

The fact that adjectives cannot be used as prepositions and that prepositions cannot be used as adjectives as in (25b) and (25e), respectively, may be attributable to the restriction on underspecification in (23), that either N or V, but not both, be underspecified. This does not explain, however, why adjectives 12 The question of whether there is a basic syntactic category for a given lexical item is only briefly addressed, not because it is not an important question, but rather, because the topic in itself deserves a whole paper. While linguists may have intuitions on the basic category of a multifunctional lexical item on the basis of e.g. historical derivations, it is not clear whether naive native speakers know about the basic categorial properties of a multifunctional lexical item. 13 [Since I wrote this chapter I became aware of cases of prepositions being used as verbs: e.g. 'to up', 'to down' (Anne-Sophie Bally pc). This constitutes a counter example to the claim that prepositions cannot become verbs. More cases of this type need to be documented.]

170

ISSUES IN THE STUDY OF PIDGIN AND CREOLE LANGUAGES

cannot be used as verbs nor prepositions as nouns or as verbs. It thus appears that directionality is involved in the account of multifunctionality. Why should this be so? Could it be because, as is shown in (21), adjectives and prepositions are, featurewise, mixed categories? Indeed, while adjectives share their [+N] feature with nouns and their [+V] feature with verbs, prepositions share their [-N] feature with verbs and their [-V] feature with nouns. If the mixed character of adjectives and prepositions is responsible for the complementary distribution between the possible combinations in (24) and the impossible ones in (25), then why should it be so? If, on the contrary, the mixed character of adjectives and prepositions is irrelevant to the differential distribution in (24) and (25), then we want to know what the cause of this complementary distribution is. In any case, the discrepancy between (24) and (25) suggests that underspecification is not a multidirectional phenomenon. A sound explanation of this fact is called for. Another question is whether there is a difference between underspecification and derivation by zero affixation in the account of cases of morphological conversion. The problem is the following. In (23), for each combination of [aN, ßv], either [αN] or [ßv] but not both, is underspecified for the value + or -, yielding the multicategorial possibilities schematised in (24). Now, let us suppose that both variables (a and ß) are underspecified for the value + or -. In this case, all the combinations in (26) obtain. (26)

[aN, ßv]:

[+N,-V](N) [+N, +V] (A) [-N, -V] (P) [-N, +V] (V)

Among these possibilities, we find nouns used as verbs ([+N, -V]/[-N, +V]), and verbs used as nouns ([-N, +V]/[+N, -V]). The fact that nouns can be used as verbs (and verbs as nouns) in the languages of the world is well documented. This phenomenon has been referred to in the literature as morphological conversion.14 Lieber (1992) has argued that predictable cases of morphological conversion15 involve the affixation of a morpheme that has all the properties of an affix, except that it lacks a phonological representation. The null affix determines the syntactic category of the derived words as well as their conceptual structure. The category switches shown in (26) would thus, in some cases, yield the same results as zero affixation. Are there arguments that permit us to choose

14 Morphological conversion in Haitian is documented in Brousseau, Filipovich and Lefebvre (1989), Fauchois (1983), Filipovich (1987), Hall (1966), Sylvain (1936), Tinelli (1970). 15 Unpredictable cases of morphological conversion constitute new lexical entries that are created by the process of relisting rather than by a regular derivational process.

MULTIFUNCTIONALITY AND THE CONCEPT OF LEXICAL ENTRY

171

between the two possible analyses of cases of morphological conversion, that is, underspecification of both N and V for categorial features, or zero affixation? The underspecification proposal in (26) would also allow adjectives, defined as [+N, +V], to be used as prepositions, defined as [-N, -V], and vice versa. To the best of my knowledge, such cases of morphological conversion have not been reported. On an underspecification account of the type in (26), such cases cannot be ruled out without recourse to a special proviso. Zero affixation, however, produces only the possible cases. This is because phonologically null affixes, like other affixes, are specified for categorial features. In current analyses, the categorial features of the morphological head define those of the derived word. In this view, a null affix specified for the features [-N, +V] selects a nominal base and the derived word is a verb. Likewise, a null affix specified for the features [+N, -V] selects a verbal base and the derived word is a noun. If there are no phonologically null affixes bearing the features [+N, +V] or [-N, -V], the impossible derivations will not be produced. The implication of such a conclusion is that cases of multifunctionality related through underspecification could differ in only one major feature.16 In this view, cases of multifunctionality of major category lexical items would be accounted for in either of two ways: by categorial underspecification involving a change in one major feature or by zero affixation involving a complete categorial switch. We are left with the following problem: while nouns can become verbs and verbs can become nouns, adjectives and prepositions do not participate in categorial switch in the same way. Why? Again, is it related to the fact that, featurewise, they are mixed categories? As can be seen from the above discussion, the topic of the categorial underspecification of major category lexical items is far from being exhausted. In addition to further developing the questions raised in this section, future research should explore the problem of categorial underspecification in light of recent models such as those developed in Hale and Keyser (1993) or in Marantz (1997). Whether these new models would allow for a unified account of the phenomena under discussion is a question for future research. Regardless of the model chosen, however, the same facts need to be accounted for. 6.4.3 The underspecification of functional category lexical items There are also multifunctional functional items. Their multifunctionality has also been accounted for in several instances in terms of underspecification. A few examples are provided below.

16 A similar constraint has been proposed in Lefebvre and Muysken (1988) in order to restrict the phenomenon of transcategoriality.

172

ISSUES IN THE STUDY OF PIDGIN AND CREOLE LANGUAGES

Ritter (1992) provides a theoretically motivated account of the syntactic properties of Haitian yo occurring as a third-person plural personal pronoun and the plural marker, as in (18). She proposes that the plural marker is the head of a syntactic projection Num(ber)P. She argues that, while first and second-person pronouns are D(eterminer)s, third-person pronouns are derived by raising Num to D, as is schematically represented in (27).

On this account then, both the plural marker and the third-person plural pronoun belong to the category Num(ber). The multifunctionality of other functional items has been accounted for in recent literature by saying that the same functional item may occur as the head of different syntactic projections. For example, in order to account for the various functions of English -ing, Travis (1992) (building on Abney 1987 and Valois 1991) proposes that -ing may occupy various syntactic positions. For example, in (28a), -ing is the head of INFL(ection); in (28b), it is the head of DP (determiner phrase); in (28c) it is the head of ASPP (aspectual phrase); and in (28d), it is also the head of ASPP, but in this case, it is also linked to the head of DP by the feature [+ -ing]. (28) a. b. c. d.

I approve of Mary singing the song. I approve of Mary's singing the song. John's fixing of the sink was skillful. John's fixing of the sink was surprising.

ENGLISH (from Travis 1992)

The Haitian and Fongbe definite determiners constitute a case of multifunctionality. Indeed, in both languages, the same definite determiner that occurs in nominal structures also plays a central role in clause structures (Lefebvre 1998b; Lefebvre and Massam 1988). Furthermore, as is documented in detail in Lefebvre (1998b), when the definite determiner occurs in the clause, there are scopal ambiguities with respect to the interpretation of the determiner. In Lefebvre (1998b) (building on Lefebvre and Massam 1988), it is proposed that there is only one definite determiner: in all of its functions, the definite

MULTIFUNCTIONALITY AND THE CONCEPT OF LEXICAL ENTRY

173

determiner bears the features [+ definite] and [+ anaphoric]. This accounts for the semantic unity of this lexical item in its various functions. It is further proposed that the definite determiner is underspecified in categorial features. This allows it to head different syntactic projections. While in nominal structures, the determiner is the head of DP, in clause structures, it can head different functional projections such as CP, MOODP, TP and ASPP. This analysis accounts for the scopal ambiguities that are manifested when the definite determiner occurs in the clause structure. So far, we have seen cases of major category lexical items underspecified for one major feature and cases of functional lexical items not specified for categorial features. We still have to consider cases of lexical items that can be used either as major category items or functional category items. A case in point is the lexical item meaning 'to say', which can be used either as a verb or a complementiser, as we saw in (8) and (9) for Saramaccan and Sranan, respectively. Similar lexical items are widespread in West African languages (see e.g. Lord 1976). Tardif (2000) offers a unified analysis of 'to say' in Fongbe. In the next section, I summarise his analysis. An under specification account of Fongbe 'to say9: verb and complementiser According to Tardif (2000) (see also the references cited therein), the Fongbe lexical item may be used as a verb meaning 'to say', as in (29a), the second verb of a series, also meaning 'to say', as in (29b), or a THAT-type complementiser, as in (29c).17 6.5.

(29)

a.

FONGBE

person 2sg say word give 'Who did you talk to?' b.

FONGBE

Koku take story say to 'Koku told a story to Asiba.' c

(from Anonymous 1983: VIII, 2) Asiba (=(46a) in Kinyalolo 1993b) FONGBE

3sg order COMP 2sg NEG come INS 'He ordered that you do not come.' (from Anonymous 1983: VI, 3) 17 On the basis of Ewe data, Lord (1976) argues that the complementiser function of the verb meaning 'to say' is the result of the reanalysis of this verb as a complementiser from its occurrence as the second verb of a series. Lord claims that her analysis extends to other West African languages, including Fongbe. On the basis of Fongbe data, Kinyalolo (1993b) takes issue with Lord (1976) and claims that is a lexical verb in all the contexts in which it occurs. Tardif (2000) evaluates the arguments of the aforementioned analyses. He convincingly concludes that has the double function of a verb and a complementiser.

174

ISSUES IN THE STUDY OF PIDGIN AND CREOLE LANGUAGES

Building on the literature on the relationship between the verb meaning 'to say' in English and the complementiser 'that' (see e.g. Partee 1973), on the semantics of complementisers (see e.g. Bresnan 1979), and on the semantic underspecification of the verbs meaning 'to say' in world languages (see e.g. is underspecified both Amberber 1996), Tardif (2000) proposes that Fongbe semantically and categorially.18 He adopts the general proposal in Amberber (1996: 8) according to which the verb meaning 'to say', across languages, is underspecified for the semantic feature 'inner manner', unlike the other verbs of its class, such as shriek, yell, mumble etc., which are all specified for this semantic feature. According to Amberber, this accounts for the properties of the verbs meaning 'to say' that depart from those of the other verbs in the same class. For example, while the other verbs may appear without a complement (e.g. John mumbled/yelled), say has to appear with something (e.g. *John said/ok John said this/something!that...), which is the manifestation of 'inner manner' for this verb. On Tardifs analysis, is semantically defined only as [+realis], a feature that determines its semantic properties both as a verb and a complementiser. In its function as a complementiser, it is in complementary distribution with the [-realis] complementisers. It is also in a semantic relationship with the mood markers occurring between the subject and the verb. The form is also categorially underspecified, bearing only the feature [+V]. This feature determines its syntactic properties. First, it can only occur as the head of VP or CP. The form is not marked for any value of N; this explains why it is never used as a noun or as an adjective, both bearing the feature [+N]. On the basis of a theory-internal argument, Tardif (2000) argues that cannot be specified for the feature [-N] either. Recall from (21) that verbs and prepositions share the feature [-N]. In Chomsky's (1981) framework, verbs and prepositions assign a case to their object; thus, they are case assigners. The generalisation is made that [-N] lexical items are case assigners, whereas [+N] lexical items are case assignees. Tardif argues that the verb is not a case assigner; therefore, it cannot bear the feature [-N]. It follows that cannot be used as a preposition, a lexical item that bears the feature [-N]. On Tardif's proposal, the sole two features [+ realis] and [+V] can account for all the

18 In many languages of the world (e.g. Gbe, etc.), the verb meaning 'to say' is also used as a complementiser. It is also the case that, in many languages of the world (e.g. English, Quechua, etc.), a deictic term meaning 'that' is used as a complementiser. Languages seem to choose either one of these two options. Sranan, however, appears to have both taki from English talk and datie from Dutch dat (see Plag 1993). This suggests that, at some deep level of representation, there might be a semantic relationship between 'to say' and 'that'. What this relationship is remains a question for future research. The question of why particular languages choose one option rather than the other is also a question for future research.

MULTIFUNCTIONALITY AND THE CONCEPT OF LEXICAL ENTRY

175

occurrences of Fongbe behaving sometimes like a verb and sometimes like a complementiser. Such case studies show that multifunctional lexical items need not depart from the monosemy principle and may be registered as single lexical entries, provided that they are semantically and categorially underspecified. It goes without saying that finding the right semantic and categorial specification for multifunctional lexical items is not an easy task. Furthermore semantic and categorial underspecification are not independent from each other. The challenge is to harmonise them. More analyses of this type are needed to deepen our understanding of multifunctionality. 6.6,

On the relationship between multifunctionality and grammaticalis ation Grammaticalisation is a particular type of reanalysis. Reanalysis is a process by which a particular form that signals one lexical entry becomes the signal of another lexical entry (see e.g. Lightfoot 1979). Through reanalysis, a phonological form that signals a major category lexical entry may become the signal of another major category lexical entry (e.g. V and P, as in (13c)) or it may become the signal of a functional category (e.g. V and COMP, as in (8) and (9)). In the literature, the latter cases of reanalysis are referred to as cases of desemanticisation or grammaticalisation (see e.g. Heine, Claudi and Hünnemeyer 1991; Heine and Reh 1984; Hopper and Traugott 1993; Traugott and Heine (eds) 1991). To the best of my knowledge, there are no abrupt cases of grammaticalisation (see e.g. Traugott and Heine 1991: l) 1 9 . All documented cases of ongoing grammaticalisation involve gradual changes, in the sense that, during some time span, a phonological form may have two functions: the original one and the new one (see e.g. Heine and Reh 1984: 57; Mufwene 1996a; Sankoff 1980, 1991). The two functions associated with a given form may coexist for a long period of time or even stabilise as such. If it appears that grammaticalisation (like other cases of reanalysis) creates multifunctional lexical items such as Fongbe it does not appear that it creates a lexical entry that is totally independent from the original one. Grammaticalisation may thus be seen as the acquisition of a new function by a given form, rather than as the creation of another lexical entry. 19 A potential counterexample to this generalisation is the grammaticalisation of the numeral meaning 'one' as an indefinite determiner in some creole languages. For example, Bruyn (1996: 39) argues that the Sranan indefinite singular determiner wan (from English 'one') did not undergo changes between the 18th century and the 20th century. She concludes that the emergence of the indefinite determiner must have taken place in the early stage of Sranan as a rare case of more or less instantaneous grammaticalisation.

176

ISSUES IN THE STUDY OF PIDGIN AND CREOLE LANGUAGES

Can the relationship between multifunctionality and grammaticalisation be characterised within the same terms? In this view, the process of grammati calisation could be described as changing a fully specified lexical entry into an underspecified one, allowing the grammaticalised form to fulfil both the original and the new functions. The process of reanalysis or grammaticalisation viewed in this way is represented in (30) for the Haitian verb bay/ba 'to give', proposed to have been reanalysed as a preposition meaning 'for' (see (4) and (5)). (30) a.

Original lexical entry bay/ba 'to give' [-N, +V]

b.

Revised lexical entry bay/ba 'to give/for' [-N, ßv]

HAITIAN

In (30), V is not specified for any value. If ß is assigned a positive value, bay/ba will be realised as a verb. If ß is assigned a negative value, bay/ba will be realised as a preposition. I leave to future research the task of establishing an underspecified LCS that will allow the reanalysed lexical entry in (30) to be interpreted either as a verb meaning 'to give' or a benefactive preposition meaning 'for'. We just saw that grammaticalisation creates multifunctional lexical items. A logical question to ask is whether all cases of multifunctionality necessarily arise through grammaticalisation. This is the topic of the next section. 6 . 7 . Grammaticalisation and multifunctionality are not synonyms It is not rare in the literature on creoles to see cases of multifunctionality being referred to as cases of grammaticalisation. In the paragraphs that follow, I discuss the necessity of arguing that a specific lexical item has undergone gram maticalisation, for multifunctionality and grammaticalisation are not synonyms. With current literature (cited earlier), I will assume that grammaticalisation is a process of linguistic change that takes place over time. I will also assume that grammaticalisation consists in assigning a phonological form to a lexical entry that has no phonological representation, by the reanalysis of a form that is the phonological representation of another lexical entry. Consistent with these assumptions is the presupposition that grammaticalisation can be observed over time. First, a phonologically null form can be shown to acquire a label over time: while at some point in time, a syntactic position can be shown to be phonologically null, at some other point in time, it can be shown to be phonologically overt. Second, the phonological representation of the newly filled position can be argued to also fill another position at times. The original and the new positions manifest a difference in word order. A case in point is the grammaticalisation of an adverb as a TMA marker. For example, in Tok Pisin, the sentence-initial adverbial baimbai 'later, afterwards' has been reanalysed as

MULTIFUNCTIONALITY AND THE CONCEPT OF LEXICAL ENTRY

177

an aspectual marker (see Sankoff 1991; Sankoff and Laberge 1973). In its adverbial function, baimbai occurs sentence-initially; as an aspectual marker, it occurs in its reduced form between the subject and the verb, as is shown in (31). (31)

a. Baimbai ologeta wok bilong yumi, later all work of us i ken kamap. PR

TOK PISIN

FUT.POSSIB grow

'Later, all our work will (be able to) flourish.'(=(2) in Sankoff 1991) b. Mi bai kisim. TOKPISIN I FUT get 'I'll get it.' (=(9) in Sankoff 1991) Likewise, Heine and Reh (1984: 120) report that in Bari, a West African language, the adverb dé 'then, afterwards' became a future or irrealis preverbal marker. As has been observed by Spagnolo (1933: 105-106), in its adverbial function, dé occurs sentence-initially; as a future marker, it occurs between the subject and the verb, as is shown in (32). (32)

a.

Dé

nàn

then I do 'Then I do... '

b.

Nàn

dé BARI

I FUT do T shall do...' (from Heine and Reh 1984: 120)

The examples in (31) and (32) show that, at one point in time, the position corresponding to the TMA marker in the (b) sentences between the subject and the verb had no phonological representation. At another point in time, the adverb of posteriority was reanalysed/grammaticalised and became the phonological form of a marker filling the initially empty position. As a TMA marker, the form occurs between the subject and the verb rather than clause-initially. The claim that there has been grammaticalisation in these two cases is thus supported by linguistic arguments. A problem in the literature on creole languages is that grammaticalisation is often not argued for, but simply assumed on the basis of the synchronic multifunctionality of a given lexical item. A case in point is the lexical item táki/táa used as a verb and a complementiser in Saramaccan, as discussed in Veenstra (1996a). At the beginning of a chapter entitled "Serial verb constructions: grammaticalisation", Veenstra (1996a: 153) warns the reader: "This discussion is based on synchronic data only". A few pages later, he writes "I will now present syntactic evidence to argue that in the latter use táa has been grammaticalised and acquired the status of complementiser" (p. 156). His arguments consist of two syntactic tests showing that the verb táa and the complementiser táa have different syntactic properties. Veenstra concludes with a

178

ISSUES IN THE STUDY OF PIDGIN AND CREOLE LANGUAGES

hypothesised grammaticalisation path whereby the serial verb táa has been reanalysed as a complementiser. While the tests presented by the author clearly argue for the multifunctional status of táa in Saramaccan, they do not show that it was grammaticalisation that led to this situation. The methodology used in Veenstra (1996a, 1996b) is similar to that in Byrne (1987: 154) who, on the basis of synchronic syntactic tests, concludes that táa has been reanalysed as a complementiser: "Finally, it is evident from the extraction pattern and THAT-trace effects that many Saramaka have reanalysed táa in predicate adjective contexts as an actual complementiser". While the multifunctionality of Saramaccan taki/táa has clearly been established, the evidence that taki/táa has undergone grammaticalisation in Saramaccan still remains to be shown. In Sranan, the form taki also has the functions of a verb and a complementiser. The methodology in Plag (1993, 1995) argues for the grammaticalisation of taki in this language. On the basis of early texts, Plag reconstructs a first stage where complement clauses are introduced by a phonologically null complementiser. He reconstructs a second stage where taki is used between speech-act verbs and complement clauses. More recent stages are also reconstructed, showing all the contexts where the complementiser taki ended up being used over time. In addition to the historical arguments, the double function of taki in contemporaneous Sranan is documented on the basis of syntactic tests (e.g. as a verb, taki may participate in the predicate cleft construction, but, as a complementiser, it cannot (see (11)). It thus appears that the methodology in Plag is an adequate one for arguing that the Sranan complementiser taki has evolved through grammaticalisation. Authors should be careful not to take arguments of synchronic multifunctionality as arguments for grammaticalisation because, if grammaticalisation always produces multifunctional lexical items, multifunctionality need not be exclusively the product of grammaticalisation (see below). Furthermore, arguments for grammaticalisation as a language-internal phenomenon are most important in the case of creole languages, for they bear on the crucial question of their genesis and development. In recent literature on the development of creoles, the question of apparent cases of grammaticalisation has been raised. I now turn to the discussion of this issue. 6 . 8 . Apparent cases of grammaticalisation in creole languages Ordinary cases of language-internal grammaticalisation have been argued to occur in creole languages as well as in other languages. The reanalysis of the Tok Pisin presentential adverb baimbai as a pre-verbal marker is a case in point (see(31)). 20 20

For a discussion of such cases, see e.g. Lefebvre (1998a: 41-46), and the references therein.

MULTIFUNCTIONALITY AND THE CONCEPT OF LEXICAL ENTRY

179

Aside from these ordinary cases of grammaticalisation, there are cases that Bruyn (1996: 42) refers to as apparent cases of grammaticalisation, "where a feature does not result from grammaticalisation that took place within the creole language itself but rather from the transfer of the result of a process of grammaticalisation that has taken place in another language". Bruyn exemplifies her claim by showing that the double function of taki as a verb and a complementiser in Sranan may be attributable to the substratum lexicons (most probably Gbe lexicons). She begins her argument by taking issue with Plag's analysis, cited in the preceding section. While it may be attractive to trace instances of grammaticalisation in creole languages, we should be cautious about overemphasising internal grammaticalisation and losing sight of external influences. According to Plag (1993, 1994b), the complementiser function of Sranan taki derives from a mainly language-internal grammaticalisation process. There are, however, reasons for assuming that there has been less internal development in this respect between the 18th and the 20th century than Plag claims. The use of taki as an unambiguous complementiser, i.e. with a verb other than those expressing speech acts, is indeed not frequent in the 18th century. However, there are a few instances that do not fit the development sketched by Plag, including the following case from Van Dyk (ca. 1765): (11)

oenno zabi takki 2pl know say/that ben pree lange PAST play with 'You know that no one has months.'

no wan zomma NEG one person da drom na negi moen tem. the drum at nine month time been playing on that drum for nine (Van Dyk ca. 1765:108)

Thus, the use of taki as a complementiser with cognitive verbs appears to have been possible at an earlier stage than the mid-19th century, contrary to what Plag (1993) concludes. Apart from the few cases were taki is a genuine complementiser, as in (11), its use as a connective element between speech act verbs and their complements is very frequent in the 18th century as well as in later stages of Sranan. The way in which it is used resembles that of its counterparts in certain West African languages, for example in the intonation break after taki, apparent from the use of a comma, especially in the Schumann texts. Even when there has been a language-internal expansion of function to some extent, it appears unwarranted to assign only a minor role to the relevant West African languages [...]. A possible contribution of West African languages is acknowledged by Plag (1993, 1994b), but he focusses on the grammaticalisation of taki as a language-internal development nevertheless. In my view, this gives an oversimplified picture of the development of taki in Sranan.

So, although Plag may have used an adequate methodology to his point, as we saw in the previous section, the additional facts brought out by Bruyn21 21 See also her footnote 11, p. 41.

180

ISSUES IN THE STUDY OF PIDGIN AND CREOLE LANGUAGES

cast doubt on the conclusion that the complementiser function of taki is a result of a language-internal development. Rather, it would appear that the properties of taki in Sranan replicate those of corresponding lexical items in the substratum languages. Likewise, it is not unlikely that the properties of taki in Saramaccan replicate those of corresponding lexical items in the substratum languages. In fact, at first glance, it seems that taki, in both Sranan and Saramaccan, has properties that are quite similar to those of Fongbe22 discussed earlier. Now that extensive data on in Fongbe (see Kinyalolo 1993a, 1993b; Tardif 2000), and on taki in Sranan and Saramaccan (see the references cited above) are available, it has become possible to do a detailed comparison of taki in these two creoles, and of in Fongbe. The results of such a comparison should help settle the issue. If, in addition to the historical data provided by Bruyn cited above, it turned out that taki in Sranan and in Fongbe had the same properties, we would have to conclude that the complementiser function of taki is in fact only an apparent case of grammaticalisation, following Bruyn's claim. A similar conclusion could hold true for Saramaccan also, against Veenstra's analysis reported above.23 Such a conclusion would not come as a surprise, for there are other similar cases discussed in Lefebvre (1998a) based on Haitian Creole and its substratum languages. Similar cases are also reported by Keesing (1991: 334) based on Melanesian Pidgin and its substratum languages. As Keesing puts it: 'The point of all this is that a Melanesian Pidgin grammatical element derived from an English lexical form may have acquired this grammatical force not through the usual chain of grammaticalisation but through a direct calquing onto a substrate pattern". Thus, Bruyn's suggestion that, in creole languages, there are apparent cases of grammaticalisation that are attributable to transfer from another language where grammaticalisation has taken place is borne out by data that are independent from those on which her claim was based. 6.9.

The transfer of substratum multifunctional lexical items into a creole From the above discussion, it appears that what is in fact being transferred into the creole is a lexical item with all of its functions, thus a multifunctional lexical entry. In Keesing's (1991) terms, the transfer is the result of calquing. I would say that it is the result of the relexification of an underspecified lexical entry. Lexical entries are lexical entries whether they are fully specified or not.

22 Gbe languages, among which Fongbe, are argued to be among the major substratum languages of Saramaccan (Smith 1987), and of Sranan (Arends 1995). 23 For a preliminary discussion of the similarities between Fongbe and Saramaccan, see Muysken (1994b).

CHAPTER 7

ON THE SEMANTIC OPACITY OF CREOLE LANGUAGES* The idea that creole languages are more transparent than languages not known to be creoles, has been in the literature for some time. While working on the details of the lexicon, morphology, syntax and semantics of creole languages, and more particularly on Haitian Creole, I often had the intuition that, somehow, the creole was as opaque as—and in some cases, even more opaque than—its source languages. Intuitions do not constitute arguments for or against any proposal or hypothesis (any more than impressions do). Nevertheless, they may provide the researcher with the curiosity to investigate a particular question further. In this column (my fourth and final one), I evaluate the Semantic Transparency Hypothesis, as formulated in Wekker (1982), Seuren and Wekker (1986), and, more recently, in McWhorter (1998a). According to Seuren and Wekker (1986: 62), the Semantic Transparency Hypothesis "stipulates a strict one-to-one correlation between semantic elements and structural surface units (morphemes)". Semantic transparency "limits learning and computing efforts to a minimum" (p. 68). Semantic transparency is thus postulated as a means of easing acquisition. As Seuren and Wekker put it: "Intuitively speaking, ST [semantic transparency] can be seen as a property of surface structures enabling listeners to carry out semantic interpretation with the least possible machinery and with the least possible requirements regarding language learning" (p. 64). The Semantic Transparency Hypothesis makes predictions about the form of creole languages. First, it predicts a lack of semantic extensions and idiosyncrasies in the lexicon. Second, it predicts 'lack of morphology' as a feature of creole languages. Seuren and Wekker (1986: 62) justify this point in the following way: "...we feel that there is sufficient ground for putting forward the claim that the lack of morphology is a specific consequence of a more general feature of creole languages, their property, that is, of maximising semantic The content of this chapter was originally published as my fourth and last column for the Journal of Pidgin and Creole Languages (Lefebvre 2001c). The research underlying this chapter has been funded by SSHRCC. I would like to thank Marie-Claude Labbé, Renée Lambert and Olivier Tardif for their assistance in researching the data on which this chapter is based. I would like to thank the following people for their most useful comments on a previous version: Anne-Marie Brousseau, Robert Papen, Ingo Plag, Jean-François Prunet, Jeff Siegel. Finally, many thanks to Andrée Bélanger for formatting the manuscript.

182

ISSUES IN THE STUDY OF PIDGIN AND CREOLE LANGUAGES

transparency". Haitian Creole constitutes a direct counterexample to this claim, for it has a number of productive affixes, ten or eleven, depending on the analysis, as discussed in Brousseau, Filipovich and Lefebvre (1989), in Filipovich (1987), in Lefebvre (1998a); on DeGraff's (2001) analysis, Haitian Creole has even more affixes. McWhorter's (1998a: 797) proposal allows for derivational affixes in creole languages; in his view, however, "in languages known as creoles, derivation is generally semantically transparent".1 His assertion is congruent with the view that semantic drift creates endless idiosyncratic lexicalisations over time. Since creole languages are relatively young languages, there has not been time for idiosyncratic lexicalisations to be formed in these languages. McWhorter's approach thus predicts that a creole derivational affixes will be semantically transparent. Third, the Semantic Transparency Hypothesis predicts the absence of rules modifying basic word order, such as predicate raising, so as to limit data processing (Seuren and Wekker 1986: 65). Fourth, on the Semantic Transparency Hypothesis, the semantic interpretation of utterances should proceed directly from S-structure: "the amount of processing needed to get from semantic analyses to surface structures and vice versa, is kept to a minimum" (Seuren and Wekker 1986: 67). Finally, the transparency strategy predicts the least recourse possible to language-specific phenomena and the most recourse possible to languageuniversal phenomena (Seuren and Wekker 1986: 66). In this column, I evaluate the predictions of the Semantic Transparency Hypothesis against data drawn from Haitian Creole. The Semantic Transparency Hypothesis is tested by two key questions. First, considering the various components of the grammar (the lexicon, the morphology, the syntax and the interpretive component), is it the case that there is a one to one correspondence between forms and meaning at different levels of representation? Second, if creoles are transparent, we would expect non-transparent features of their contributing languages to be eliminated in creolisation. Is this in fact the case? The answer to the first question constitutes the core of this chapter. The second question will be addressed in the conclusion. The chapter is organised in the following way: Data pertaining to the various components of the grammar will be discussed in turn. I begin with the lexicon and the morphology. The discussion of the syntactic component of the grammar (e.g. basic word order and movement rules), and of the interpretive component will follow. The issue of language-specific versus languageuniversal phenomena in creole languages will then be briefly addressed. My first 1 Note that part of McWhorter's (1998a) proposal is formulated with respect to the phonological component of the grammar, and more specifically the role of tones, not discussed in this chapter.

ON THE SEMANTIC OPACITY OF CREOLE LANGUAGES

183

conclusion is that there are several ways in which opacity may manifest itself in a creole, at all levels of representation. The data I present in this chapter support the view that creole languages are no more transparent than languages not known to be creoles, and that in some cases, they are even more opaque than these languages. My second conclusion, discussed as part of the section entitled Global evaluation of the Semantic Transparency Hypothesis, is that nontransparent features of substratum languages are not necessarily suppressed in creolisation, since a lot of idiosyncrasies of the creoles' substratum languages are transferred into the creoles in creole genesis. 7.1. Lexicon The Semantic Transparency Hypothesis predicts a one-to-one correspondence between semantic units and surface morphemes. It is not compatible with semantic idiosyncrasies, the presence of idiomatic expressions, some paradigmatic facts involving functional categories, lexical entries that have no overt phonological form, nor with cases of categorial neutralisation (or even cases of multifunctionality). In fact, the Semantic Transparency Hypothesis is not compatible with any kind of idiosyncrasy. This section argues that the Hypothesis is not supported by the Haitian data, since the Haitian lexicon is full of idiosyncrasies. 7.1.1 Lexical semantics A one-to-one correspondence between semantic units and surface morphemes is not compatible with the fact that there are creole lexical items which have more than one specific meaning, as is exemplified in (l)-(7). (1)

ansasinen

'to murder, to abuse, to aggress against; to mutilate (e.g. one's leg); to trash (e.g. a tree)' (from Lefebvre 1999a: 69-70)

(2)

betize

'to joke; to deceive; to work in vain, to accomplish nothing'

(from Valdman et al. 1981) (from Valdman et al. 1981)

(3)

kase

'to break; to slim down'

(4)

kraze

(5)

pike

'to break to pieces, to break by spreading; to disperse' (from Lefebvre 1999a: 73; from Valdman et al. 1981) 'to swoop, to pounce; to string, to prick, to stab; to transplant (seedlings); to draw (a card); to kill (by magical means)' (from Valdman et al. 1981)

184

ISSUES IN THE STUDY OF PIDGIN AND CREOLE LANGUAGES

(6)

ranje

'to arrange, to fix; to increase; to poison' (from Valdman et al 1981)

(7)

taye

'to cut, to trim; to make love to; to whip' (from Valdman et al 1981)

Each lexical entry in (l)-(7) presents an idiosyncratic interpretation. For example, it is not clear how the meanings 'to draw (a card)' and 'to kill (by magical means)' in (5), can be built into the semantic representation of the Haitian verb pike, which otherwise has several other meanings. Likewise, the meaning 'to poison', associated with the verb ranje, in (6), is not easily describable in terms that would make it derivable from the same semantic representation as 'to arrange, to fix', and so on, and so forth. These idiosyncratic interpretations constitute counterexamples to the Semantic Transparency Hypothesis. 7.1.2 Idiomatic expressions By definition, the meaning of idiomatic expressions does not correspond to their literal interpretation. Thus, idiomatic expressions are idiosyncratic. Idiomatic expressions are not compatible with the Semantic Transparency Hypothesis either. Nonetheless, they abound in Haitian Creole. A few examples are provided in (8) and (9). (8)

HAITIAN

a. b. c. d. e. f.

bay chènn manje kòb kase pye bat vont monje dan chèche wout

g. prannanm (9)

GLOSS

give chain eat money break foot hit stomach eat tooth look for road take soul

'to wind (e.g. a clock)' 'to spend' 'to stumble' 'to worm secrets out of' 'to grind one's teeth' 'to look for alternative medical treatment' 'to charm, bewitch' (from Valdman et cd. 1981)

Nouvèl la kase bra news DEF break arm T had bad news.'

m. me

HAITIAN (from Védrine 1992)

Given the fact that idiomatic expressions are idiosyncratic, their presence in a creole constitutes a counterexample to the Semantic Transparency Hypothesis. (Further examples are provided in Lefebvre (1999a) and in Lumsden (1999b).)

ON THE SEMANTIC OPACITY OF CREOLE LANGUAGES

185

7.1.3 Paradigms of functional categories Although a great deal of the Haitian substratum languages' functional categories have been reproduced in the creole, some paradigms of functional categories contain fewer items than the corresponding paradigms in the substratum languages (for an extensive discussion of this point, see Lefebvre 1998a and chapter 3). For example, in Fongbe there is a paradigm of clause final functional items that gives the point of view of the speaker on the proposition. This paradigm includes a negative marker and a marker of insistence. While the marker of insistence has been reproduced in Haitian Creole, the negative marker has not (see Lefebvre 1998a: 213-217). Thus, this particular Haitian paradigm may be said to not be as rich as the corresponding Fongbe one. It could be argued that a syntactic position that can host more than one lexical item, as in this particular Fongbe case, has presumably more visibility than a position that can host only one lexical item, as in this particular Haitian case. It could further be argued that, from the learner's point of view, the creole data are more opaque in this particular case, than the substratum ones. As such, they constitute a counterexample to the Semantic Transparency Hypothesis. 7.1.4 Phonologically null forms In Haitian Creole, there are a number of lexical entries that have a phonologically null form. Such lexical entries are those that have a grammatical function without having an overt phonological representation (for an extensive discussion of this point, see Lefebvre 1998a, see also chapter 3). For example, Lumsden (1989, 1991) states that the argument following the head noun in the Haitian nominal structure must be marked for case, by virtue of the Case Filter (see Chomsky 1981), in spite of the fact that this argument is not accompanied by an overt case marker. He further argues that this phonologically null case has the properties of a genitive case (e.g. 's in English) rather than that of an objective case (e.g. of in English). By Lumsden's analysis, then, in (10), the argument following the head noun is followed by a phonologically null genitive case. (10)

timoun

Man

child

Mary GEN

ø

a

HAITIAN

DEF

'(the) Mary's child' The least we can say about phonologically null forms is that, if anything, they are not transparent, for their properties have to be deduced from those of the phrases in which they 'covertly appear'. Thus, from the language learner's point of view, this situation is more opaque than what we find in a language like, say English or Fongbe, where there are overt case markers. For those who do not like the idea of phonologically null forms, the point still holds. As was pointed

186

ISSUES IN THE STUDY OF PIDGIN AND CREOLE LANGUAGES

out to me by Ingo Plag (p.c.) and shown in (10), in Haitian there is no overt grammatical expression of possession. In a transparent system, we would expect overt expression of possession. 7.1.5 Multifunctionality or cases of categorial neutralisation A lexical item that fulfils more than one grammatical function is said to be multifunctional. The presence of multifunctional lexical items in creole languages could be viewed as violating the Semantic Transparency Hypothesis that "stipulates a strict one-to-one correlation between semantic elements and structural surface units (morphemes)" (Seuren and Wekker 1986: 62). It is a well documented fact that creole languages, like other languages, present a fair amount of lexical items that have more than one function. For example, Haitian Creole presents a variety of multifunctional lexical items (see e.g. Lefebvre 1998b, 2001b; Védrine 1992). In Lefebvre (2001b) [see chapter 6], it is shown that multifunctional lexical items in Haitian Creole can be divided into three groups: one in which a given lexical item may belong to more than one major syntactic category (i.e. noun, verb, adjective, preposition), one in which a given lexical item may serve both as a major syntactic category and as a functional category, and another one in which a given lexical item may belong to more than one functional category.2 In these cases, it is not always possible to establish a one-to-one correspondence between morphemes and semantic units. Cases of multifunctionality may thus be seen as opaque as opposed to transparent. Furthermore, from the learner's point of view, multifunctional items should be more difficult to learn since their various meanings and functions ought to be deduced from the context in which they occur. 7.1.6 Summary The Haitian Creole lexicon includes several types of data that constitute counterexamples to the Semantic Transparency Hypothesis. Most of the semantic extensions in (l)-(7) are idiosyncratic. They are thus incompatible with the Semantic Transparency Hypothesis. Likewise, the idiomatic expressions in (8) and (9) are idiosyncratic by definition. They are thus not compatible with the Semantic Transparency Hypothesis. Some paradigms also present some idiosyncrasies. Phonologically null lexical entries, as in (10), are inherently less transparent than phonologically overt ones. Multifunctional lexical items may also be seen as being less transparent than non-multifunctional lexical items. The Haitian data reported on in this section thus constitute counterexamples to the Semantic Transparency Hypothesis with respect to the properties of the creole 2 Extensive references to multifunctional lexical items in other creole languages are also provided in Lefebvre (2001b).

ON THE SEMANTIC OPACITY OF CREOLE LANGUAGES

187

lexicon. The facts presented in this section argue in favour of another view, according to which creole lexicons would present several facets of semantic opacity. Seuren and Wekker (1986: 67) also mention that the Semantic Transpa rency Hypothesis implies that secondary constructions, such as comparative grammaticalised constructions, "will seldom be found in creole languages". It is a well known fact that comparative constructions in the Caribbean creole langua ges make use of lexical items meaning 'to surpass' to encode comparison rather than using a grammaticalised form (see e.g. Jansen, Koopman and Muysken 1978). In my view, this is not necessarily an implication of the Semantic Transparency Hypothesis; rather, it is due to the fact that creole languages tend to reproduce the semantic and syntactic properties of the lexical entries of their substratum languages, as is extensively discussed in Lefebvre (1998a) and in chapter 3. For example, while Fongbe uses the verb hú 'to surpass' in comparative structures of superiority, as is shown in (11), Haitian uses the verb pase 'to surpass' to express the same meaning, as is illustrated in (12). (11)

kló

hú

Àsíbá

Koku be.big surpass Asiba 'Koku is bigger than Asiba.' (12)

Jan gwo pase John be.big surpass Mary 'John is bigger than Mary.'

FONGBE

(=(2) in Platt 1992) Mari

HAITIAN

The Haitian data in (12) parallel the Fongbe ones in (11); they do not reproduce the French structure in (13). (13)

Jean est plus gros que Marie John is more big than Mary 'John is bigger than Mary.'

FRENCH

It could be claimed that particular substrate structures are retained in the creole because they are semantically transparent. The question of whether the structures in (11) and (12) are more transparent than the one in (13) could be debated on the basis of a list of criteria, a task which I leave for future discussion. The fact is, however, that there are opaque substrate features that have been reproduced in the creole, as is extensively discussed in Lefebvre (1998a), and in the concluding section of this chapter.

188

ISSUES IN THE STUDY OF PIDGIN AND CREOLE LANGUAGES

7.2. Morphology This section addresses the question of whether the productive affixes of Haitian are semantically transparent. Due to limitation of space, only one subset of affixes will be examined: suffixes that refer to a place of origin, the prefix ti- and the prefix de-. 7.2.1 Suffixes that refer to a place of origin Haitian has two suffixes used to form nouns that refer to a place of origin: -wa and -yen. Examples are given in (14). (14) a. b.

Tchad-wa Soudan-n-wa Petyonvil-wa Ougand-yen Pakistan-n-yen Jakmèl-yen

'from Chad' HAITIAN 'from Soudan' 'from Petionville' 'from Uganda' HAITIAN 'from Pakistan' 'from Jacmel' (from Brousseau p.c. cited in Lefebvre 1998a: 310)

The principles governing the choice between the two affixes -wa and -yen (either semantic or phonological), if any, are not known. However, as was pointed out by Brousseau (p.c.), native speakers of Haitian know exactly which of the two affixes to use when presented with a large list of place names. The Semantic Transparency Hypothesis would predict that there should be only one affix to derive nouns that refer to a place of origin in Haitian Creole. Nonetheless, Haitian Creole has two different affixes, encoding the same notion. 7.2.2 The prefix ti Another example involves the prefix ti-. In Haitian, there is a noun which has a full form, pitit 'child' (Valdman et al. 1981), or a reduced form ti 'kid' (Valdman 1996). The short form also serves as a prefix that attaches to nominal bases. Its basic meaning with animate bases is 'child of x', as is illustrated in (15) and (16). (15)

(16)

chat chen towo

'cat' 'dog' 'bull'

ti-chat ti-chen ti-towo

'kitten' 'puppy' 'calf'

chwal

'horse'

ti-chwal

'colt'

HAITIAN (from Valdman et al. 1981) HAITIAN (from Filipovich 1987: 34)

The prefix ti- also has the extended meaning 'a small x' (see e.g. Filipovich 1987: 34-35; Sylvain 1936: 32; Valdman et al. 1981). As a diminutive prefix, ti-

ON THE SEMANTIC OPACITY OF CREOLE LANGUAGES

189

may occur either with inanimate nominal bases, as in (17), or with animate nominal bases, as in (18). (17) a.

flè chante

'flower' 'song'

b. wòch 'stone' kay 'house' pyebwa 'tree' c. kiyè van

(18) nèg moun bet fi gason medam mesye

'spoon' 'wind'

ti-flè ti-chante

'small 'small song'

flower'

HAITIAN

(from Sylvain 1936: 32) ti-wòch 'pebble' HAITIAN ti-kay 'small house' ti-pyebwa 'bush' (from Filipovich 1987: 34-35) ti-kiyè 'teaspoon' HAITIAN ti-van 'breeze' (from Valdman et al 1981)

'man' ti-nèg 'young man' HAITIAN 'person' ti-moun 'child' 'animal' ti-bèt 'insect, bug' 'woman' ti-fi 'girl' 'male (man, boy), waiter' ti-gason 'boy' 'women, ladies' ti-medam 'young ladies, girls' 'Messrs., men' ti-mesye 'young men' (from Valdman et al 1981; Valdman 1996)

Semantic extensions of words containing the prefix ti- are manifested in euphemisms such as in (19), or in expressions such as those in (20). The data in (19) and (20) constitute idiosyncrasies. (19) ti-bezwen ti-devan ti-wòz ti-gigit ti-pijon (20) ti-bourik ti-chat ti-fi ti-gran moun ti-grann ti-malis ti-nouris ti-tèt

(bezwen 'need') (devan 'front') (yvdz 'rose') (gigit 'penis') (pijon 'pigeon')

'genitals (of a small boy)' HAITIAN 'genitals' 'menstrual period' 'penis' 'penis' (from Valdman et al 1981; Valdman 1996)

(bourik 'donkey') 'rude person' HAITIAN (chat 'genitals (female)') 'honey (sweetheart)' (fi 'woman') 'virgin' (gran moun 'adult, old person') 'precocious child' (grann 'grandmother') 'precocious little girl' (Malis 'folk character 'canny person' known for his cleverness') (nouris 'wet nurse, 'new mother' nursing mother') (tèt 'head') 'birdbrain' (from Valdman et al 1981; Valdman 1996)

190

ISSUES IN THE STUDY OF PIDGIN AND CREOLE LANGUAGES

The data on the semantics of ti- do not meet the semantic transparency criterion. Indeed, there is semantic extension associated with ti-, from 'child of x' , as in (15) and (16), to 'small x', as in (17) and (18), and there are expressions that make use of this prefix that constitute idiosyncrasies, as in (19) and (20). 7.2.3 The prefix deThe Haitian prefix de- is not as semantically transparent as the Semantic Transparency Hypothesis would predict. As will be shown below, de- selects either attested or non-attested verbal bases. In the first case, it is either exclusively interpreted as an inversive prefix, or it is semantically ambiguous between an inversive and a privative interpretation. In the second case, it is always assigned a privative interpretation. There are also cases where the prefix de- is semantically vacuous. As will be shown below, words containing de- also present other types of idiosyncrasies. The Haitian prefix de- selects verbs. It derives new verbs which all have a meaning opposite to that of the base form. The examples in (21), all taken from Valdman et al. (1981), are built on the same model: a verb (e.g. chaje 'to load'), preceded by the prefix de-, is assigned a meaning that is opposite to that denoted by the verb without the prefix (e.g. de-chaje 'to unload'). (21) debare 'to clean (remove obstacles)' debouche 'to uncork, to open (bottle)' deboure 'to remove (stuffing)' deboutonnen 'to unbutton' dechaje 'to unload, to ejaculate' de fèt 'to undo, to disconcert, to get rid of' defòme 'to deform' degonfle 'to deflate, to scatter, to make room for' degrese 'to lose weight, to remove fat' degwosi 'to smooth out, to lose weight, to rough-hew'

bare HAITIAN 'to block off, to fence in' bouche 'to stop up, to cork' boure 'to stuff' boutonnen 'to button' chaje 'to load' fè 'to do, to make, to produce' fòme 'to form' gonfle 'to blow up' grese 'to grease, to oil' gwosi 'to grow larger, to put on weight'

ON THE SEMANTIC OPACITY OF CREOLE LANGUAGES dekachte 'to unseal' dekloure 'to remove nails' dekochte 'to unhook' dekole 'to unstick' dekonnekte 'to disconnect, to cut off (isolate)' dekontwole 'to fluster'

hackte 'to seal' kloure 'to nail' 'to hook (a door)' kole 'to stick (glue)' konnekte 'to plug in'

'to control, to supervise, to verify' dekouvri kouvri 'to uncover, to clear up (weather)' 'to cover' delase lase 'to untie, to unwind' 'to lace' demakdnen makònen 'to untangle, to disentangle' 'to tangle, to tie together' demele mele 'to untangle, to get along (manage), 'to mix (to associate), to shuffle (cards)' to mix (blend)' deplase plase 'to move' 'to live together (commonlaw marriage), to give a daughter to someone in common law marrige' ploge deploge 'to unplug' 'to plug in' deploye pliye 'to unfold, to unroll' 'to bend, to fold' deplötonnen plötonnen 'to unwind' 'to wrap, to wind' derespekte respekte 'to respect' 'to be disrespectful' desele sele 'to unsaddle' 'to saddle' sere de sere 'to tighten' 'to losen' sèvi desèvi 'to clear (the table)' 'to use, to serve (be fo use), to associate with, to serve, to serve (sorship)' antere detere 'to bury' 'to unearth, to dig up' vise devise 'to screw' 'to unscrew'

191

192

ISSUES IN THE STUDY OF PIDGIN AND CREOLE LANGUAGES

devlope 'to unwrap, to develop (grow)' dewoule 'to unroll' dezabiye 'to undress' dezanflè 'to become less swollen' dezipe 'to unzip' dezobeyi 'to desobey'

vlope 'to wrap, to fold' woule 'to roll, to roll up' abiye 'to dress, to clother (provide clothing)' anfle 'to swell' zipe 'to zip' obeyi 'to obey' (from Valdman et al 1981)

In all the examples above, de- may be assigned an inversive meaning, and in some cases, it is the only meaning it can be assigned. For example deboutonnen can only be assigned the inversive interpretation 'to unbutton' of boutonnen 'to button'. It cannot be assigned a privative interpretation, for deboutonnen 'to unbutton' cannot be interpreted as 'to remove the buttons from the clothes'. However, several examples in (21) are ambiguous between an inversive and a privative interpretation. For example, while in one of its uses, bare means 'to block off' (Valdman et al 1981), debare means 'to clean (remove obstacles)'. The verb debare can be interpreted as having a meaning opposite to bare, in which case it is interpreted as inversive. It can also be interpreted as involving the notion of 'removing an obstacle' in which case it has a privative interpretation. Likewise, the verb degrese can be interpreted as having a meaning opposite to grese 'to oil, to grease' (see Valdman et al. 1981); it can also be assigned the privative interpretation 'to lose weight'.3 Several other words containing the prefix de- also present this kind of semantic ambiguity, between an inversive and a privative interpretation. As is pointed out in Brousseau, Filipovich and Lefebvre (1989) and Filipovich (1987), the Haitian prefix de- may also attach to a base that is unattested as an actual verb. For example, while there is a noun branch 'branch', there is no verb *branche which would mean 'to branch a tree (i.e. to attach branches to a tree)'. However, there is a verb de-branche 'to prune (e.g. a tree)' which appears to have been derived from the unattested verb *branche 'to branch'. Several verbs containing the prefix de- are derived on this model. A sample of these verbs are listed in (22). Note that in all these cases, de-V can only be assigned a privative interpretation. For example, while one can conceive

3

The semantics of this pair of lexical items is further discussed in Chaudenson (1996a: 27).

ON THE SEMANTIC OPACITY OF CREOLE LANGUAGES

193

of pruning a tree, the reverse notion, that is, that of attaching branches to a tree, does not correspond to anything in the real world. (22) a.

b.

debagaje debouse debranche dechouke

'to move (change residence)' HAITIAN 'to spend (money)' 'to prune' 'to uproot, to get someone fired, to remove from office, to plow' degagannen 'to cut the throat of, to overcharge' degouse 'to shell, to husk' dekreta 'to remove the crest' dekwennen 'to skin' demachwele 'to smash one's face in' depaye 'to take apart a woven straw object' deplancheye 'to remove the floor boards' derasine 'to uproot' devantre 'to disembowel' dezose 'to debone' (from Valdman et al. 1981) demwèle 'to remove the marrow' HAITIAN deplezi 'to displease' (from Filipovich 1987:42)

The nouns from which the verbs in (22) are derived are listed in (23) together with the corresponding unattested denominal verb. (23) a.

b.

bagay bous branch-bwa chouk gagann gous krèt kwenn machwè zo pay planche rasin vant mwèl plezi

'thing' *bagaje HAITIAN 'purse' *bouse 'branch' *branche 'stump' *chouke 'throat' *gaganne 'pod, clove' *gouse 'cockcomb' *kreta4 'rind' *kwennen 'jaw' *machwèle 'bone' *zose 'straw, chaff' *paye 'wooden floor' *plancheye 'root' *rasine 'belly, stomach' *vantre (from Valdman et al 1981) 'marrow' *mwèle HAITIAN 'pleasure' *plezi5 (from Filipovich 1987: 42)

Finally, it is not clear whether de- should be analysed as a prefix in debaba 'to mow (a lawn), to weed' (from Valdman et al. 1981). Not only is there no verb 4 Also discussed in DeGraff (2001). 5 The Haitian verb meaning 'to please' is plè.

194

ISSUES IN THE STUDY OF PIDGIN AND CREOLE LANGUAGES

*baba, but there is no noun *baba with a related meaning. The lexical entry baba in Valdman et al. (1981) is glossed as 'idiot'. It thus appears that baba and debaba are not morphologically related.6 So far, we have seen that the prefix de- may attach to attested (see (21)) and to non-attested (see (22) and (23)) verbal bases. We have also seen that, in the first case (see (21)), Haitian words derived by means of the prefix de-, may be assigned an inversive interpretation exclusively, or they may be assigned an interpretation that is either inversive or privative. In the second case (see (22)), the Haitian words derived by means of the prefix de- can only be assigned a privative interpretation. Finally, it is not clear whether debaba and baba are morphologically related. In my view, these facts are not entirely clear. Additional data further complicate the picture. For example, according to Valdman et al. (1981), the verb demele means 'to untangle', 'to get along (manage)' and 'to shuffle (cards)'. The action of shuffling cards has the effect of mixing the cards rather than untangling them. This is quite an idiosyncrasy. Another set of data concerning the Haitian prefix de- is presented in (24). In this case, the prefix de- attaches to a verbal base, derived from a noun. For example, the noun plim 'feather, body hair, fur' is converted into a verb by means of the suffix -e yielding plimen 'to pluck'. This verb may be prefixed by de-, yielding de-plimen which also means 'to pluck'. Thus, with or without the prefix de-, the verb has the same privative interpretation. In Valdman's et al (1981) dictionary, the pairs of verbs in (24) are given the same definition.7 (24) degrennen grennen < grenn HAITIAN 'to shell' 'to shell' 'seed' dekale kale < kal 'to chip off, to peel off' 'to shell, to peel' 'covering(bark, skin, shell)' deplimen plimen < plim 'to pluck' 'to pluck' 'feather, body hair, fur' derefize refize < refiz 'to refuse' 'to refuse' 'refusal' devire vire < vire 'to turn' 'to turn' 'turn' (from Valdman et al. 1981)

6 DeGraff (2001) also provides a discussion of this word. 7 DeGraff (2001) considers as emphatic the verb derefize as opposed to refize (see (24)). He also considers as emphatic demegri, as opposed to megri, both meaning 'to lose weight'. To my knowledge, no other author provides these emphatic interpretations for some words containing the prefix de-. Furthermore, an emphatic interpretation of the other verbs in (24) (e.g. dekale 'to chip off, to peel off', deplime 'to pluck', etc.), is hard to imagine.

ON THE SEMANTIC OPACITY OF CREOLE LANGUAGES

195

The data in (25) present a similar pattern of pairs of verbs. In this case, however, there are no nouns involved in the derivation of the verbs containing the prefix de-. (25) a.

b.

dechire, chire 'to tear, to rip' HAITIAN dechifonnen, chifonnen 'to wrinkle, to rumple' degengole, gengole 'to rush down' (from Valdman et al 1981) degobye, gobye 'to belch' HAITIAN (from Filipovich 1987: 41)

Some pairs of adjectives also follow this pattern, as is examplified in (26). (26)

demefyan, mefyan

'mistrustful'

HAITIAN (from Valdman et al 1981)

In (24), (25) and (26), the prefix de- is semanticalfy vacuous. The existence of pairs of synonyms, such as those in (24), (25) and (26), is not compatible with the view that the morphology of creole languages is semantically transparent. Indeed, the prefix de- is sometimes interpreted as inversive (see (21)), sometimes as privative (see (22)), sometimes as either inversive or privative (see (21)); in some cases, it is even semantically vacuous (see (24)-(26)). From the learner's point of view, this cannot be considered as semantically transparent. Another point has to do with the semantic extensions found with verbs containing the prefix de-. Examples are given in (27). (27) degwosi dechouke

'to lose weight, to smooth out, to rough-hew' HAITIAN 'to get someone fired, to remove from office, to uproot, to plow' dechaje 'to unload, to ejaculate' degagannen 'to cut the throat of, to overcharge' degonfle 'to deflate, to scatter, to make room for' degrade 'to demote, to damage, to take away the supernatural power of a dead person' degrennen 'to shell, to break a bill, to knock out, to thin out (forest)' dekoupe 'to eat copiously, to cut up, to interrupt' dekreta 'to cut of the comb of a rooster, to mistreat' demanbre 'to batter (a person), to ravage, to weaken, woodoo dwelling left by the ancestors' detripe 'to gut, to defame' (from Valdman et al 1981)

196

ISSUES IN THE STUDY OF PIDGIN AND CREOLE LANGUAGES

The semantic extensions of the kind found in (27) are not compatible with the view that the morphology of creole languages is transparent. 7.2.4 Summary The data pertaining to a subset of the derivational affixes of Haitian Creole do not appear to support the Semantic Transparency Hypothesis, as there are various ways in which Haitian Creole morphology may be considered to be opaque rather than transparent. The examples presented so far are the following. There are two suffixes that refer to a place of origin; although they are phonologically distinct, they mean exactly the same thing. They appear to be allomorphs of an unusual kind, because the selection of one or the other is not contextually predictable. This is a first type of manifestation of opacity. A second type of opacity is offered by the prefix ti-. This prefix also presents semantic extensions and some of the words in which it is found have idiosyncratic meanings. Finally, the prefix de- presents other types of opacity. It can be assigned two interpretations (inversive and privative), but it can also be semantically vacuous. In some cases the meaning of the words derived with deis not predictable (for an extensive discussion of Haitian de-, see also DeGraff 2001). [See also chapter 10 for further discussion of the Semantic Transparency Hypothesis as related to the morphology of Haitian Creole.] 7 . 3 . Syntax: basic word order and movement rules According to Seuren and Wekker (1986: 65), the Semantic Transparency Hypothesis predicts "a uniform strategy for arranging verbs and their arguments (subject, object, indirect object). Thus, rules that bring about variations in the order of subject (S), verb (V), object (O), and though less crucially, indirect object (IO), will be untypical of creole languages". As part of the Semantic Transparency Hypothesis, the uniformity strategy would preclude predicate raising, and presumably other movement rules, because they would modify the basic word order of the language. In Seuren and Wekker's view, movement rules would also violate the simplicity criterion that "implies that the amount of processing needed to get from semantic analysis to surface structures, and vice versa, is kept to a minimum (p. 67)". It is the case that there is no predicate raising (that is, verb movement to INFL) in Haitian Creole (see e.g. DeGraff 1992b). However, in my analysis, this rule is absent from the grammar of Haitian Creole due to parametric options that are pertinent to both creole and non-creole languages (see Chomsky 1981 and Pollock 1989, and related literature), and that have nothing to do with the "semantic transparency", the "uniformity strategy", or the "simplicity criterion". As is extensively discussed in Lefebvre (1998a: 351-355, and the references therein), predicate raising is available in languages that have verbal inflectional

ON THE SEMANTIC OPACITY OF CREOLE LANGUAGES

197

morphology, and it is not available in languages that have no verbal inflectional morphology. Since Haitian does not have the relevant verbal inflectional morphology, predicate raising is not available in this language. Having dispensed with this apparent problem, I now turn to the discussion of movement rules8 that alter the basic word order of Haitian Creole. The data reported on below will be shown to be incompatible with the prediction in Seuren and Wekker (1986). In addition to Wh-movement involved in questions, clefts, topicalisations and relative clause formation (see e.g. Koopman 1982a, 1982b; Piou 1982a, 1982b), Haitian Creole exhibits some other movement rules. Argument raising is one of them. For example, the argument of the verb rete 'to remain, to be left over' may appear in two surface positions, as is shown in (28). In (28a), the argument follows the verb and the subject position is optionally filled with the expletive pronoun li. In (28b), the argument appears in the subject position and the basic position is empty. (28) a. b.

Li rete yon sèpan (nan pannye an). HAITIAN it remain a snake in basket DEF 'There remains a snake (in the basket).' (=(14) in Dumais 1988) [Yon sèpan]i rete [e]i nan pannye an HAITIAN a snake remain in basket DEF 'A snake remains in the basket.' (=(15) in Dumais 1988)

The Haitian verb manke 'to be missing, to lack' also allows for argument raising. In this case, either the object or the locative argument can be moved to the subject position of the verb. This is shown in (29). (29) a. b. c.

Li manke sèl nan soup la. it be.missing salt in soup DEF 'There lacks salt in the soup.' Sèl manke nan soup la. salt be.missing in soup DEF 'Salt is lacking from the soup.' Soup la manke sèl. soup DEF be.missing salt 'The soup lacks salt.'

HAITIAN (=(16) in Dumais 1988) HAITIAN (=(17) in Dumais 1988) HAITIAN (=(18) in Dumais 1988)

Sentences containing the Haitian verb sanble 'to seem, to resemble'9 may have two surface realisations. In the first one, the subject of the embedded clause is 8 I am using 'mouvement rule' in a non-technical sense here. 9 For extensive theoretical discussions of the properties of the verb sanble 'to seem', see e.g. DeGraff (1992a, 1992b), Déprez (1992a), Dumais (1988), Law (1991, 1992), Lumsden (1993), Massam (1989), Ritter (1991), etc. For competing views on whether this Haitian construction

198

ISSUES IN THE STUDY OF PIDGIN AND CREOLE LANGUAGES

realised in the embedded clause and the subject position of the matrix verb is optionally filled by the expletive pronoun li. In the second, the subject of the embedded clause occurs in the subject position of the matrix verb and the embedded subject position is filled by the resumptive pronoun li. This is shown in (30). (30) a. b.

Li samble Jan te malad. HAITIAN it seem John ANT sick 'It seems that John has been sick.' (=(13) in Dumais 1988) Jan sanble li te malad. HAITIAN John seem RES ANT sick 'John seems to have been sick.' [Lit.: 'John seems as if he had been sick.'] 10 (=(12) in Dumais 1988)

More data on subject raising in Haitian are presented in Massam (1989). In addition to these argument raising rules that modify the basic word order, Haitian exhibits the double object construction. In this construction, the so-called indirect object precedes the so-called direct object, as is shown in (31). (31)

Mwen bay I montre Pòl liv la. HAITIAN I give / show Paul book the 'I gave/showed Paul the book.' (=(19) and (25b) in Lumsden 1994)

The word order in (31) departs from the more general word order where the direct object immediately follows the verb. In Haitian there appears to be several contexts where the basic word order may be altered. These are predicted to be 'untypical' of creole languages, and yet they do exist in Haitian Creole. 7 . 4 . Interpretive facts The intepretive component of the grammar is another area where the Semantic Transparency Hypothesis can be tested. In this section, I look at the interpretation of the tense, mood and aspect subsystem of the grammar. I consider the interpretation of clauses involving argument alternations; I examine verb doubling phenomena and the interpretation of the predicate cleft construction. The one-to-one correspondence between surface forms and

really involves movement, see e.g. Law (1991, 1992) and Ritter (1991). For the implications of these data for Government theory, see Déprez (1992a) and Law (1991, 1992). 10 See Lappin (1984) for a discussion of sentences such as John seems as if/like he is unhappy.

ON THE SEMANTIC OPACITY OF CREOLE LANGUAGES

199

semantic interpretations postulated by the Semantic Transparency Hypothesis is not borne out by the data presented below. 7.4.1 The inte rp retation of tense, mood and aspect The interpretation of the subsystem of tense, mood and aspect of creole languages is not as transparent as it may seem at first glance. A detailed discussion of this subsystem of the grammar for Haitian may be found in Lefebvre (1996a: 231-311). Here, I only provide a few examples showing some non-tranparent aspects of this subsystem of the grammar. Haitian allows for bare sentences, that is, sentences in which there is no preverbal marker. The temporal interpretation of bare sentences is computed from the various components that participate in establishing the aspectual properties of a clause (see e.g. Déchaine 1991). These components include the aspectual class of the verb (see Lumsden 1995a), the specificity of the direct object of the verb (see Damoiseau 1988), and the specificity of the subject (see Bentolila 1987). For example, abare sentence containing a dynamic verb with a specific object is interpreted as present perfect; a sentence containing a dynamic verb with a non-specific object is ambiguous between a general present and a past interpretation (see Lefebvre 1996a: 253). A bare sentence containing a resultative verb is ambiguous between a present and a present perfect interpretation (see Lefebvre 1996a: 204; Lumsden 1995a). Finally, a bare sentence containing a stative verb is always interpreted as present (see Damoiseau 1988; Déchaine 1991; Lumsden 1995a). A sentence containing the marker of anteriority te is either interpreted as past or as pluperfect depending on the aspectual class of the verb (see Damoiseau 1988; Lumsden 1995a). A clause containing the definite future marker ap, or the indefinite future marker a-va may be assigned future or future perfect interpretation (see e.g. Lefebvre 1996a: 243; Sylvain 1936: 87). The interpretation of ap as a future marker or as an imperfective marker is also contingent upon the aspectual class of the verb (see Lefebvre 1996a: 288; Lumsden 1995a). Furthermore, as is shown in Lefebvre (1996a: 288-291), the presence of a negative marker within a clause may alter the basic interpretation of the irrealis TMA markers. The various points mentioned above constitute counterexamples to the Semantic Transparency Hypothesis. They show that the Haitian TMA system certainly does not correspond to Bickerton's (1984) more classic one. 11 In my

11 Bickerton's "classic" system has a reduced inventory of TMA markers and a reduced inventory of compound tenses, as compared to that reported on in Lefebvre (1996a). Furthermore, Bickerton presents none of the details reported on here, based on Lefebvre (1996a).

200

ISSUES IN THE STUDY OF PIDGIN AND CREOLE LANGUAGES

view, the data mentioned above constitute examples of opacity in the Haitian TMA system. 7.4.2 The interpretation of clauses involving argument alternations A clause containing a verb allowing for argument alternations is semantically ambiguous. For example, the S-structure in (32) may be assigned an inchoative, a verbal passive or an adjectival passive interpretation. (32) a. b. c.

Branch la kase. branch DEF break Inchoative The branch broke.' Verbal passive The branch was broken (by someone).' Adjectival passive The branch is broken.'

HAITIAN

Several examples of this type are reported in Massam (1989). The fact that the Haitian structure in (32) may be assigned several interpretations (see (32a, b, c)) violates the Semantic Transparency Hypothesis. Interestingly enough, in this case, the creole is more opaque than its contributing languages, because it lacks the lexical material that disambiguates between the above structures in French and in West African languages, respectively. As is shown in (33), the three interpretations of (32) are morphologically distinguished in French by the selection of different auxiliaries. (33) a.

b.

c.

Inchoative La branche a cassé. DEF branch HAVE broken The branch broke.' Verbal passive La branche a été cassée. DEF branch HAVE BE broken The branch was broken (by someone).' Adjectival passive La branche est cassée. DEF branch BE broken The branch is broken.'

FRENCH

FRENCH

FRENCH

As is shown in (34), the three Haitian interpretations in (32), are also morphologically distinguished in Fongbe. The sentences in (34b) and (34c) are distinguished by nyí 'to be' and ɖò 'to be at', respectively. Both are distinguished from (34a) in the fact that, while they require a 'be' verb, (34a) does not [and it would preclude it].

ON THE SEMANTIC OPACITY OF CREOLE LANGUAGES (34) a.

b.

c.

Inchoative Kófú 6 gbà. cup DEF break The cup broke.' Verbal passive Kófú 6 nyí gbìgbà. cup DEF be broken 'The cup was broken (by someone).' Adjectival passive Kófú ó ò gbìgbà. cup DEF be. at broken The cup is broken.'

201 FONGBE

FONGBE

FONGBE

The French data in (33), and the Fongbe data in (34), are thus more transparent than the corresponding Haitian data in (32), which are highly ambiguous. 7.4.3 Verb doubling phenomena Clauses containing what looks like a copy of the predicate involve four constructions, three of which are discussed in this section (see below for the fourth): temporal adverbial, as in (35), causal adverbial, as in (36), and factive clauses, as in (37). In the examples below, the first occurrence of the verb is an exact replica of the second. The examples in (35)-(37) provide the Fongbe and the Haitian data in parallel. This is to show that the Haitian structures, once again, are more opaque than the corresponding Fongbe ones, due to the lack of overt markers in Haitian. For example, the Haitian structures in (35) and (36) lack a lexical item corresponding to Fongbe tróló in (35), and to Fongbe útú in (36), in such a way that we find the same Haitian structure in (35) and (36) with different interpretations. Thus the same Haitian structure may be assigned either a temporal or a causal adverbial interpretation. Likewise, from (37), it can be seen that while Fongbe has an overt nominal operator, Haitian has a phonologically null one. (35) Temporal adverbial Wá Jan wá {tróló) bò Mari yì. FONGBE Rive Jon rive (epi) Mari pati. HAITIAN arrive John arrive as.soon.as and Mary leave 'As soon as John arrived, Mary left.' (=(1) in Lefebvre 1994b) (36) Causal adverbial Wá Jan wá útú Mari yi. Rive Jan rive Mari pati. arrive John arrive cause Mary leave 'Because John arrived, Mary left.'

FONGBE HAITIAN (=(2) in Lefebvre 1994b)

202

ISSUES IN THE STUDY OF PIDGIN AND CREOLE LANGUAGES

(37) Factive Wá é-è Jan wá 6, víví FONGBE Rive ø Jan rive a, fè HAITIAN arrive OP John arrive DEF make(-happy) nú no ton manman li kòntan for mother his happy 'The fact that John arrived made his mother happy.' (=(3) in Lefebvre 1994b) Once again, the Haitian structures are more ambiguous, therefore more opaque, than the Fongbe ones. 7.4.4 The interpretation of cleft constructions The predicate cleft construction is the fourth structure involved in verb doubling phenomena. The construction is illustrated in (38). Note from the translation that the interpretation of the structure involves a contrastive reading on the predicate of the clause. (38) Predicate cleft Wá w Jan wá. Se rive Jan rive. it.is arrive it.is John arrive Tt is arrived that John did.' (not e.g. leave)

FONGBE HAITIAN (=(4) in Lefebvre 1994b)

As is extensively discussed in Larson and Lefebvre (1991) and in Lefebvre (1990), in this construction, the clefted constituent may be assigned an intepretation that goes beyond what is actually found in the clefted phrase. For example, the clefted constituent in (39) may be assigned three different contrastive interpretations: one bearing on V, another one bearing on the VP, and a last one bearing on the internal argument of the verb. (39) Se monje Jan monje pen an HAITIAN that.is eat John eat bread DEF Tt is EATING the bread that John did.' (not e.g. throw it away) Tt is EATING THE BREAD that John did.' (not e.g. wash the dishes) Tt is eating THE BREAD that John did.' (not e.g. eat the apple) (=(44) in Lefebvre 1990) The example in (40) presents similar focal ambiguities.

ON THE SEMANTIC OPACITY OF CREOLE LANGUAGES

203

(40) Se mache Jan mache al lekol. HAITIAN it.is walk John walk to school 'It is WALK that John did to school.' (not e.g. run) 'It is WALK TO SCHOOL that John did.' (not e.g. run home) 'It is TO SCHOOL that John walked.' (not e.g. to the park) (=(23) in Larson and Lefebvre 1991: 251) Other similar examples are discussed in Lefebvre (1990) and Larson and Lefebvre (1991). Finally, when the affected argument of a verb has been clefted, the contrastive interpretation of the cleft constituent bears either on the noun phrase, or on the whole VP. This is illustrated in (41). (41) Se pen an Jan manje. HAITIAN it.is bread DEF John eat 'It is the BREAD that John ate.' (not e.g. the apple) 'It is EATING THE BREAD that John did.' (not e.g. wash the dishes) (=(53) in Lefebvre 1990) These interpretive facts are not semantically transparent in the sense defined in the introduction to this chapter. They require interpretive rules, that do far more than establishing a one-to-one correspondence between the S-structures and their interpretations. 7.4.5 Summary The subsets of data presented in this section illustrate various facets of the opacity found in the interpretive component of Haitian grammar. Several constructions present cases of semantic ambiguity that are not compatible with the Semantic Transparency Hypothesis. 7 . 5 . Language specific versus language universal phenomena The last aspect of the Semantic Transparency Hypothesis is formulated in Seuren and Wekker (1986: 66) in terms of the dichotomy between language specific versus language universal features: A maximisation of semantic transparency in the sentences of a language will mean that maximal use is made in the grammar of those processes and rule systems that are least language-specific, so that the universals which leave the least room for languageparticular variation are exploited most.

Again, this claim is not supported by the Haitian data. As is extensively discussed in Lefebvre (1998a), Haitian Creole does manifest options that are considered to be marked ones. A case in point is the

204

ISSUES IN THE STUDY OF PIDGIN AND CREOLE LANGUAGES

availability of verb doubling phenomena discussed in the preceding section. As is pointed out in Koopman (1986), these structures constitute marked options because they are extremely rare among the languages of the world. Another case in point is the fact that Haitian allows for subject raising out of a tensed clause, as is shown in (30). Subject raising generally operates out of infinitival clauses. Massam (1989: 105) notes that the type of subject raising exhibited by Haitian is rare among the languages of the world; this makes it a marked option. Other marked options in Haitian Creole are discussed in Lefebvre (1998a). So, on that last point, the Semantic Transparency Hypothesis also fails. 7 . 6 . Global evaluation of the Semantic Transparency Hypothesis The Semantic Transparency Hypothesis, as is formulated in McWhorter (1998a) and in S euren and Wekker (1986), is not supported by Haitian data. In this chapter, counterexamples to the formulations of this hypothesis were shown to be found in all four major components of the grammar: the lexical, morphological, syntactic and semantic components. In all four components there are Haitian data that can be argued to show opacity rather than transparency. Types of opacity that have been reported on in this chapter are the following: First, there are cases of semantic idiosyncrasies in Haitian. These are found both in the lexicon and in the morphological component of the Haitian grammar. Second, there are a number of idiomatic expressions that, by definition, are idiosyncratic. Third, there are a number of lexical entries that have a grammatical function without having an overt phonological representation. Phonologically covert categories cannot be claimed to be transparent. Fourth, there are a number of multifunctional lexical items. These can be argued to be more opaque than lexical items that have only one function. Fifth, the derivational morphemes discussed in this chapter manifest various facets of opacity. There are two phonologically distinct morphemes encoding the notion of place of origin; these morphemes have the same semantic and syntactic properties but their distribution is not context sensitive. The prefix ti- has two meanings: 'child of x' and 'small x'; furthermore, the meaning of some words containing the prefix ti- is idiosyncratic. As for the prefix de-, in addition to manifesting the types of opacity already identified above, it was shown to exhibit another type of opacity through the fact that, in some cases, it is semantically vacuous; indeed, some pairs of verbs with or without this morpheme were shown to be synonymous (see e.g. (24), (25), (26)). Sixth, there are a number of movement rules, all involving argument movement, that modify basic Haitian word order. These rules violate the Semantic Transparency Hypothesis. Finally, the interpretation of tense, mood and aspect, the interpretation of clauses involving argument alternations, the interpretation of constructions involving verb doubling phenomena and the interpretation of cleft constructions were

ON THE SEMANTIC OPACITY OF CREOLE LANGUAGES

205

shown to not be as transparent as the Semantic Transparency Hypothesis would predict, and as presenting cases of semantic opacity that all have in common the fact that there is no one-to-one correspondence between the S-structure and the semantic interpretation of this S-structure. I consider these cases to constitute cases of opacity. On the basis of these findings, my overall evaluation is that there is opacity in Haitian Creole. It was further shown that, in some cases, Haitian Creole is more opaque than its contributing languages (see e.g. (32M34); (35)-(38)). The facts summarised above raise the following question: where do the idiosyncrasies found in the creole come from? It is my contention that a great deal of these idiosyncrasies come from the substratum languages. For example, the semantic idiosyncrasies in (l)-(7) reproduce those found in corresponding lexical entries in Fongbe (see Lefebvre 1999a: 68-78). Likewise, the idiomatic expressions in (8) and (9) constitute direct calques from Fongbe idiomatic expressions (see Lefebvre 1999a: 80-84). As is extensively discussed in Lefebvre (1998a: 303-334), the morphology of Haitian Creole is quite similar to that of its substratum languages [see also chapter 10]. As for movement rules, the argument raising phenomena discussed in (28)-(31) follow the pattern of the substratum languages rather than that of the superstratum language (French), as is shown in Lefebvre (1998a: 262-269). The idiosyncrasies found in the Haitian interpretation of tense, mood and aspect are also found in its substratum languages (see Lefebvre 1996a), as well as those found in constructions involving verb doubling phenomena and cleft constructions (see Lefebvre 1998a: 263-274). It thus appears that a great deal of the idiosyncrasies found in the creole come from its substratum languages. This suggests that at least some of the idiosyncrasies found in a creole may be simply transferred from the substratum languages in creolisation. This does not support the break in transmission hypothesis in the creation of creole languages advocated by e.g. McWhorter (1998a), or Thomason and Kaufman (1991). Furthermore, if creoles were semantically transparent, we would expect the non-transparent features of their contributing languages to be eliminated in creolisation. While it can be observed that some non-transparent features of the creole substratum languages do not make their way into the creole, as is shown, for example, in Lefebvre (1998a: 144), the examples reported on in this chapter show that non-tranparent features of substratum languages are not systematically suppressed in creolisation. In stating this, I am in no way implying that there are no innovations in creole languages. I assume that, as is the case of other languages, innovations are part of the dynamism of creole languages. Finally, Plag (2001), independently demonstrates that studies of earliest records show idiosyncrasies in the incipient creole. This argues against the claim in McWhorter (1998a: 797)

206

ISSUES IN THE STUDY OF PIDGIN AND CREOLE LANGUAGES

according to which idiosyncrasies need to develop over time, and are thus not found in young languages. Recall from the introduction that semantic transparency was postulated as a means of easing acquisition. I have shown that there is opacity in creole languages. Assuming that this is a correct characterisation of the facts, one may wonder what the implications of this state of affairs are for acquisition. I leave this question open for future discussion.

CHAPTER 8

DO CREOLE LANGUAGES REALLY FORM A TYPOLOGICAL CLASS?* McWhorter (2001) makes two very strong claims. The first one is stated in the title of his article: "The world's simplest grammars are creole grammars". The second one appears to come as a consequence of the first: "Creole grammars constitute a synchronically identifiable class". This second claim builds on McWhorter's (1998a: 790) earlier claim that creole languages constitute a "synchronically definable typological class". In this chapter, I provide an alternative way of addressing the issues of the alleged simplicity of these languages, and of the so-called similarity between them. The following three questions will be discussed in turn. First, what do creole languages really have in common? Second, why do creoles tend to be isolating? Third, why do creole languages tend to look simpler? These questions will be addressed from the point of view of the relexification account of creole genesis in Lefebvre (1998a), and the references cited therein. Section 8.4 weighs apparent simplicity against hidden complexity. Section 8.5 considers McWhorter's hypothesised creole typological features in light of the previous discussion. It will be shown that the features proposed by McWhorter as identifiying creole languages are derivable from a sound theory of how creole languages come about. Section 8.6 concludes the chapter. 8.1. What is similar among creole languages? One way of addressing the issue of the similarity between creole languages is to ask whether they form a typological class. In my view, they do not. In the * The content of this chapter was originally published under the title What you see is not always what you get: Apparent simplicity and hidden complexity in creole languages, as part of a special issue of Linguistic Typology dedicated to the question of whether creole languages form a typological class (see Lefebvre 2001d). The content of this chapter builds on a public debate on issues in creole studies between John McWhorter and myself. The debate took place March 22, 2001 at the University of California at Berkeley. I want to thank Suzan Ervin-Tripp, Dan Slobin and all the participants of this debate for their comments and insightfull questions. I want to thank Bernard Comrie, David Gil and Frans Plank for their comments on a first version of this chapter. Thanks to Andrée Bélanger who formatted it. The research underlying this chapter was financed by SSHRCC, FCAR and FIR-UQAM.

208

ISSUES IN THE STUDY OF PIDGIN AND CREOLE LANGUAGES

paragraphs that follow, I substantiate this claim on the basis of the research on creole genesis that I have been conducting over the last twenty-five years.1 When we started this research, our basic assumption was that it should be possible to account for the formation of creole languages in terms of the same processes that are at work in language genesis and language change in general; that is: relexification, a cognitive process that has been shown to play a role in the formation of mixed languages (see e.g. Media Lengua, see Muysken 1981a); reanalysis, a major process in linguistic change (see e.g. Heine and Reh 1984); and dialect levelling, a process that has been shown to take place when dialects of the same languages come into contact (see e.g. Trudgill 1986). Our hypothesis (see e.g. Lefebvre 1998 a, and the references cited therein; Lefebvre and Kaye (eds) 1986; Lefebvre and Lumsden 1989, 1994a, 1994b) was that the creators of a creole language, adult native speakers of substratum languages, use the properties of their native lexicons, the parametric values and the semantic interpretation rules of their native grammars in creating the creole. On this hypothesis, the bulk of a Creole's lexical entries is created by the process of relexification. Two other processes, fed by the output of relexification, dialect levelling and reanalysis, also play a role in the development of the creole.2 Relexification takes place in creole genesis when speakers of several substratum languages are targeting the same superstratum language (Lefebvre and Lumsden 1994a). [The process is illustrated in section 3.1.1]. In a lexicon that is in the process of being relexified, each lexical entry acquires a second phonological representation that is derived from the lexifier language. As is shown in (1), following Mous's (1994) proposal, at a certain point in the process, each lexical entry has two phonological representations.

In the history of an early creole community, at some point, the substratum languages cease to be spoken. The original phonological representations are no longer used. Consequently, they eventually disappear from the lexicon. The new lexical entries thus have the semantic and syntactic properties of the original 1

In the course of this research, I had several major collaborators whom I wish to thank for their contribution at some point or other: Anne-Marie Brousseau, Jonathan Kaye, John Lumsden, Diane Massam. 2 For an extensive discussion of these two processes and of how they interact with relexification in creole genesis, see Lefebvre (1998a), and the references cited therein, and Lefebvre (2001a) [see also chapter 9].

DO CREOLE LANGUAGES REALLY FORM A TYPOLOGICAL CLASS

209

ones, and phonological representations derived from a phonetic string in the superstratum language. This is represented in (2).

The nature of the process of relexification predicts that the creole lexical entries will have phonological representations derived from the superstratum language and syntactic and semantic properties derived from the substratum languages.3 The hypothesis was tested on Haitian Creole. The test of the hypothesis consisted in a detailed comparison of the lexicon and grammar of Haitian Creole with those of its contributing languages: French, its superstratum language, and Fongbe, one of its substratum languages.4 The details of the threeway comparison are extensively reported on in Lefebvre (1998a). The results of the linguistic test show that, to a large extent, the hypothesis is supported by the data. In the paragraphs that follow, I provide a sample of the results of this comparative study, so as to provide the reader with some background information for the discussion that follows on the issues at stake in this article. The nominal structure, the tense, mood and aspect markers, the parameters and the verb doubling phenomena will be discussed in turn. [In the original version of this chapter, this section contains a lot of details that have been deleted from the present version. These details may be found in chapter 3.] The data in (3) provide an overview of French nominal structure. They show that, in this language, the definite determiner, the possessive and the demonstrative determiners all precede the head noun, and that there can be only 3 As has been pointed out in Lefebvre and Lumsden (1994a), this account of creole genesis is a further development of the second language acquisition theory of creole genesis (see e.g. Andersen 1980; Schumann 1978; Thomason and Kaufman 1991; etc.); it is claimed that, in creole genesis involving situations where there is little access to the superstratum language, the process of relexification is used by speakers of the substratum languages as the main tool for acquiring a second language: the superstratum language. 4 Due to various constraints, we had to limit the detailed study of the substratum languages of Haitian to one language. On the basis of non-linguistic factors, Fongbe was chosen as the substratum language to be studied in detail. In no way does this methodological choice imply that Haitian is Fongbe relexified. (For a thourough discussion of the methodology of the research and the validity of the linguistic test, see Lefebvre 1998a: 52-77, and the references cited therein.)

210

ISSUES IN THE STUDY OF PIDGIN AND CREOLE LANGUAGES

one of these per noun phrase. Singular and plural forms are contrasted by (3b) showing that plural is encoded in a bound morpheme in French. (3)

a.

He

mon ce

DEF

POSS DEM crab

crabe

FRENCH

The Haitian and Fongbe nominal structures are illustrated in (4). In both languages, the determiners all follow the head noun. In both languages, a possessor phrase, a demonstrative term, the definite determiner and the plural marker may all co-occur within the same nominal structure. In both languages, the plural marker is an independent morpheme. (4)

krab [mwen 0] sa àsón [ny tòn] éló crab me GEN DEM 'these/those crabs of mine'

a yo HAITIAN 6 lέ FONGBE DEF PL (in question/that we know of) (from Lefebvre 1998a: 78)

The Haitian and Fongbe nominal structures contrast in the same way with the French nominal structure with respect to word order, co-occurrence restrictions of determiners, and with respect to whether the plural marker is a free (in Haitian and Fongbe) or a bound (in French) morpheme. [Furthermore, in Lefebvre (1998a, and the references therein), it is extensively argued that the properties of the morphemes occurring in the nominal structures of Haitian and Fongbe are extremely similar and contrast in a systematic way with those of corresponding morphemes in French. The pertinent facts are provided in chapter 3.] As is extensively argued in Lefebvre (1998a: 89 101, 2001a), the extraordinary similarity that exists between the functional categories of the Haitian and Fongbe nominal structures comes from relexification. In Haitian Creole, the verb of a finite clause is invariant. In French, however, the verb of a finite clause obligatorily bears inflectional morphology encoding tense, mood, aspect, person, and number. None of the verbal morphology found in French has made its way into Haitian. Haitian follows the pattern of its West African (non-Bantu) substratum languages in having invariant bare verbs. In both Haitian and Fongbe, temporal relationships, mood and aspect are encoded by means of markers occurring between the subject and the

DO CREOLE LANGUAGES REALLY FORM A TYPOLOGICAL CLASS

211

verb. The inventory of the TMA markers of Haitian is quite similar to that found in Fongbe. (See Bentolila 1971; Lefebvre 1996a, 1998a: 11-140). This is shown in Table 8.1. Table 8.1. The inventory of TMA markers in Haitian and in Fongbe ANTERIOR

IRREALIS

• Past/Past perfect H F te ko

• Definite future H F ap m

NON-COMPLETE • Habituai H F — nò

• Imperfective H F ap ɖò...wè

• Indefinite future H F a-va ná-wá • Subjunctive H F pou ní

(=(115) in Lefebvre 1996a: 281) Both languages have a marker which encodes anteriority. Both lexically distinguish between definite and indefinite future. The definite future markers are used to convey the speaker's attitude that the event referred to by the clause will definitely take place in the near future. By contrast, the indefinite future markers are used to convey the speaker's opinion that the event referred to by the clause might eventually or potentially take place at an undetermined point in the future. The fact that speakers of Haitian distinguish between definite and indefinite future is widely documented in the literature (e.g. Spears 1990, and the references therein; Valdman 1970, 1978). For Fongbe, this distinction is pointed out in Anonymous (1983: V, 3). Both languages have a marker glossed as "subjunctive" for convenience. This term subsumes the three meanings of pou and ní respectively: both may be interpreted as 'must', 'should' or 'may'. Both languages have a form which encodes imperfective aspect. As can be seen in Table 8.1, there is a one-to-one correspondence between the preverbal markers in the two languages, except that Fongbe has one encoding the habitual aspect, and Haitian does not. In Lefebvre (1998a: 111-140) it is argued that, while the phonological representation of the tense, mood and aspect markers of Haitian are derived from French phonetic strings, their semantic and syntactic properties follow the details of the corresponding substratum lexical entries [see also chapter 3].

212

ISSUES IN THE STUDY OF PIDGIN AND CREOLE LANGUAGES

As was stated at the beginning of this chapter, our hypothesis concerning creole genesis included the claim that the creators of a creole use the parametric options of their own grammar in assigning a value to the parameters of the language that they are creating. The hypothesis predicts that, where the parametric values of the substratum and superstratum differ, the creole should have the same parametric value as the substratum languages. With one exception (discussed below), the three-way comparison in Lefebvre (1998a: 349-374) supports this general hypothesis. As can be seen in Table 8.2, at the time the research was conducted, parameters were formulated in terms of correlations between the availability of a functional category and a related syntactic phenomenon. As has been pointed out in Lefebvre (1998a: 387), the parametric options set in the creole are the result of its creators reproducing the properties of the functional categories of their own lexicons through relexification. The correlations discussed in Lefebvre (1998a: 349-374) are summarised in Table 8.2. (For each parameter, the proposer of the correlation is mentioned within square brackets). Table 8.2. Comparison of the parametric options in the three languages under comparison Availability of (A) Verb raising to INFL (correlates with inflectional morphology on the verb) [Pollock 1989] (B) Serial verbs (correlates with lack of derivational and inflectional morphology) [Baker 1991; Muysken 1988dl (C) Double-object constructions (correlates with availability of genitive case in nominal structures) [Johnson 1991] (D) Negative quantifiers as NPs (correlates with availability of bare NPs) [Déprez 1999] (E) Verb-doubling phenomena (correlates with the properties of the determiner system) [Lefebvre 1998a: 363-374]

FONGBE

HAITTAN

-

-

+

+

+

+

FRENCH

+

" +

+

+

+

-

(=Table 13.7 in Lefebvre 1998a: 387) As can be seen in Table 8.2, the parametric options of Haitian systematically contrast with those of French and follow those of substratum languages of the type of Fongbe. There is one exception to the general pattern revealed by Table 8.2: whereas both French and Fongbe are null subject languages, Haitian is not. In recent literature, it has been proposed that languages with syntactic clitics should be considered null subject languages (see e.g. Hulk 1986; Jaeggli 1984; Roberge 1990). Both Fongbe and French have syntactic clitics, but Haitian does not (see Lefebvre 1998a: 148-157, and the references therein). Since syntactic clitics did

DO CREOLE LANGUAGES REALLY FORM A TYPOLOGICAL CLASS

213

not make their way into the creole, as will be further discussed below, the value of the null subject parameter had to be reset (see Lefebvre 1998a: 349-351). Verb-doubling phenomena involve four constructions which contain what looks like an exact copy of the predicate: temporal adverbial, as in (5), causal adverbial, as in (6), factive clauses, as in (7) and the predicate cleft construction, as in (8). (5)

Temporal adverbial Wá Jan wá (tróló) bo Màrí yi. FONGBE Rive Jon rive (epí) Mari pati. HAITIAN arrive John arrive as.soon.as and Mary leave 'As soon as John arrived, Mary left.' (=(1) in Lefebvre 1994b)

(6)

Causal adverbial Wá Jan wá útú Mari yi. Rive Jan rive Mari pati. arrive John arrive cause Mary leave 'Because John arrived, Mary left.'

(7)

Factive Wá ɖé-è Jan wá 6 víví Rive 0 Jan rive a, fè arrive OP John arrive DEF make(-happy)

FONGBE HAITIAN (=(2) in Lefebvre 1994b) FONGBE HAITIAN

nú no ton. manman li kdntan. for mother his happy 'The fact that John arrived made his mother happy.' (=(3) in Lefebvre 1994b) (8)

Predicate cleft Wá w Jan wá. Se rive Jan rive. it.is arrive it.is John arrive 'It is arrive that John did.' (not e.g. leave)

FONGBE HAITIAN (=(4) in Lefebvre 1994b)

It is a well known fact that, while verb doubling phenomena are attested in Haitian and in West African languages, they are not attested in French (see Koopman 1986; Lefebvre 1998a, and the references cited therein). Moreover, as is demonstrated in Lefebvre (1998a: 363-374), the properties of the verb doubling constructions in both Haitian and Fongbe are strikingly similar.

214

ISSUES IN THE STUDY OF PIDGIN AND CREOLE LANGUAGES

The overview of data pertaining to major subsystems of the grammar shows that Haitian Creole shares major properties with its substratum languages. Data showing that Haitian Creole lexical entries reproduce the semantic divisions of their substratum languages, in spite of the fact that their phonological representations are derived from French phonetic matrices, may be found in Lefebvre (1998a, 1999a), in Lumsden (1999a), and in the references cited in these publications. In Lefebvre (1998a: 248-301), it is shown that, to a great extent, the syntactic properties of Haitian verbs also correspond to those of the substratum languages rather than to those of French. The inventory and the properties of the Haitian derivational affixes are also argued to be extremely similar to those of the substratum languages rather than to those of the lexifier language (see Lefebvre 1998a: 303-333, and the references cited therein). The principles governing the concatenation of words into compounds in Haitian also appear to follow the substratum languages (see Brousseau 1988, 1989; Lefebvre 1998a: 334-349) [see also chapter 3]. It thus appears that the Haitian lexicon manifests the semantic and syntactic properties of its substratum languages. Similarly, Haitian reproduces the principles of concatenation and the parametric values of its substratum languages. Abstracting away from the phonological representations of the Haitian lexical entries, it appears that Haitian Creole manifests the typological features of the Gbe cluster (here represented by Fongbe) and other West African languages, the substratum languages, rather than those of French, the lexifier language. Now, if relexification has played a central role in the formation of Haitian Creole, it is logical to hypothesise that this cognitive process has also played a major role in the formation of other creole languages. By hypothesis, then, these other creoles would also reproduce the properties of their substratum languages. In his comparison of Solomons Pidgin with its source languages, English, the lexifier language, and Kwaio, an Austronesian substratum language, Keesing (1988) shows extensively that Solomons Pidgin does reproduce the properties of its Austronesian substratum languages. Keesing (1988: 1-2) writes: Sitting on a Solomon Islands mountain in 1977, reading Derek Bickerton's review article on "Pidgin and Creole Studies" (1976), I was led to think more seriously than I ever had about the history and structure of Solomon Islands Pidgin. I had earlier been struck, when I had learned Solomon Pidgin in the 1960s through the medium of Kwaio, an indigenous language I already spoke fluently, that this learning task mainly required learning Pidgin equivalents of Kwaio morphemes. The syntax of Solomon Pidgin was essentially the same as the syntax of Kwaio, although somewhat simpler and lacking some of the surface marking; in most constructions, there was a virtual morpheme-by-morpheme correspondence between Kwaio and Pidgin. (This was not

DO CREOLE LANGUAGES REALLY FORM A TYPOLOGICAL CLASS

215

just an odd local process of calquing: the Pidgin I was learning in terms of Kwaio was spoken with only minor variations throughout the southeastern and central Solomons, although it was everywhere adapted to local phonologies.) Although most of the Pidgin lexical forms were ultimately derived from English, I found this largely irrelevant to my language-learning task. The semantic categories they labelled corresponded to Kwaio ones, not English ones; grammatical morphemes corresponded to Kwaio ones, not English ones. Thus semantically Pidgin dae corresponded directly to Kwaio mae 'be dead, die, be comatose, be extinguished,' not to English "die". Pidgin baebae corresponded to the Kwaio marker of future/nonaccomplished mode, ta-, not to English "by and by".

Keesing accounts for the linguistic situation he describes in terms of calquing. That is, the substratum speakers of Solomons Pidgin calque the properties of their native languages (e.g. Kwaio) when speaking the pidgin. The type of calquing that Keesing describes corresponds to the definition of relexification given at the beginning of this section. Keesing (1988) documents the fact that calquing of the substratum properties can be observed throughout the lexicon of Solomons Pidgin. He shows that the pronominal system of this pidgin is quite similar to that of the complex system of the substratum languages in distinguishing singular, dual and plural, inclusive and exclusive first person plural, etc. He argues that the tense, mood, aspect system of Solomons Pidgin reproduces the idiosyncrasies of the system of the substratum languages. As Keesing (1988: 215) puts it: "In fact, the entire set of Kwaio particles marking the time-frame of the verb, some of which are preverbal and some postverbal, correspond in their Solomons Pidgin usage to a set of particles derived from English but carrying exactly the same import as the Kwaio particles, and placed in exactly the same slots." Keesing further shows that, as is the case in the substratum languages, Solomons Pidgin has a predicate marker. The same pattern is also found in interrogative constructions, relative clauses, etc. In short, Keesing provides extensive evidence that, while the phonological representations of Solomons Pidgin lexical entries are derived from English phonetic strings, the properties of these lexical entries do not correspond entirely to those of English lexical entries; he convincingly demonstrates that the properties of the Solomons Pidgin lexical entries do, however, correspond to those of its substratum languages, including functional category lexical entries. The following example illustrates this situation.

216 (9)

ISSUES IN THE STUDY OF PIDGIN AND CREOLE LANGUAGES a. b.

Gila la-la FP (them) FUT-SRP (they) 'They will go.' Olketa bae-i FP (them) FUT-SRP (3pl) 'They will go.'

leka. go go. go

KWAIO SOLOMONS PIDGIN (from Keesing 1988: 214)

While olketa in the pidgin derives its phonological representation from the English expression 'all together', it has the meaning and uses of the substratum strong personal pronoun gila 'them'. While bae in the pidgin derives its phonological representation from a reduced form of the English expression by and by, its meaning and uses correspond to the substratum lexical entry ta-, a future marker. As in the substratum language, the future marker of the pidgin is marked for a third person pronominal form. This pidgin form is derived from the English he, but it does not share the uses of the form it is phonologically derived from; it does, however, share the properties of the substratum forms, as can be observed by comparing the (a) and (b) sentences in (9). So, the relexification account of creole genesis predicts that Atlantic creoles will reproduce the properties of their West African substratum languages, while Pacific ones will reproduce those of their Austronesian substratum languages. Atlantic and Pacific creoles are thus expected to differ in the same areas of lexicon and grammar as West African and Austronesian languages do among themselves. For example, while the pronominal system of Solomons Pidgin reproduces the singular, dual, plural inclusive and exclusive first person plural distinctions of its substratum languages, as was mentioned above, Haitian also reproduces the particularities of its substratum languages. Consider the paradigm of personal pronouns in (10). (10) a. FRENCH

b. HAITIAN

moi T, me' toi 'you (sg)' lui/elle 'he/she/it' nous

'we/us'

C. FONGBE

mwen T, me' ou/[wu] 'you (sg)' li 'he/she/it/ him/her' nou 'we/us/ you (pl)'

vous 'you (pl)' eux/elles 'they/them' yo 'they/them' (from Valdman et al 1981)

nyè T, me' hw 'you (sg)' é(y ) 'he/she/it him/her' mi 'we/us/ you (pl)'

yé 'they/them' (=(18) in Brousseau 1995a)

While French has six forms, Haitian has only five. As in Fongbe, in Haitian, the same form serves as both first and second person plural (for further discussion, see Lefebvre 1998a: 141-143). Likewise, verb doubling phenomena of the type in (5)-(8) are only found in those creoles for which the substratum languages

DO CREOLE LANGUAGES REALLY FORM A TYPOLOGICAL CLASS

217

have them. Thus, while Atlantic creoles have these constructions, inherited from their substratum languages, Pacific creoles do not have them because thensubstratum languages do not have them. Consequently, on the relexification account of creole genesis, creole languages cannot be argued to be typologically similar. Rather, what appears to unite creoles of different geographical areas is the main process—relexification—by which they come about. In spite of this rather categorical conclusion, there is, nonetheless, one feature that creole languages appear to share: it is the fact that they tend to be isolating languages. I now turn to the discussion of this point. 8 . 2 . Why do creole languages tend to be isolating? The observation that creole languages tend to be isolating languages goes back to Hesseling (1933: xvi) and Schuchardt (1979). It is also found in Hagège (1985: 39). Mufwene (1986, 1990, 1991) shows that this tendency appears to hold true even when the contributing languages are not isolating ones. For example, he documents the fact that Kituba, a creole language that has emerged almost exclusively from contact among agglutinative Bantu languages, is an isolating language. "Kituba has selected Kikongo's seemingly marked periphrastic alternative over the more common and apparently unmarked agglutinating system" (Mufwene 1990: 12). More recently, McWhorter (1998a: 792) has proposed that lack of inflectional morphology is a feature of the creole prototype. How does the relexification account of creole genesis handle the fact that creoles tend to be isolating? The answer to this question lies in the way that functional category lexical entries acquire a label in creole genesis. According to Lefebvre and Lumsden (1994a, 1994b), this is achieved in one of two ways. First, since the creators of a radical creole do not identify as such the functional categories of the superstratum language, because they do not have enough exposure to the language, they do not relabel the functional category lexical entries of their own lexicon on the basis of those of the superstratum language; rather, they relabel them on the basis of major-category lexemes (e.g. nouns, adjectives, verbs, adverbs and prepositions) of the superstratum language. For example, the postnominal definite determiner of the Haitian substratum languages in (4) has been relabelled on the basis of a French postnominal adverb (see Lefebvre 1998a: 78-84). Likewise, the tense, mood and aspect markers of the substratum languages of Haitian in Table 8.2 have been relexified using

218

ISSUES IN THE STUDY OF PIDGIN AND CREOLE LANGUAGES

French periphrastic expressions (see Lefebvre 1996a, 1998a: 11-140) [see also chapter 3]. The absence of syntactic pronominal clitics in a creole whose contributing languages all have syntactic pronominal clitics (as is the case of Haitian, as we saw above) can also be argued to follow from this perspective. The following scenario is proposed by Brousseau (1995a). She hypothesises that the creators of Haitian relexified the clitics of their own lexicon using French strong personal pronouns. Note that these French forms were also used to relabel the lexical entries copied from the strong pronouns. So, based on this hypothesis, the copied lexical entries of all three Fongbe first person singular pronominal forms were relabelled on the basis of French moi, yielding mwen in Haitian, as is shown in (11) (where the syntactic features [± argument] stand for strong and clitic forms, respectively). (11) a. [1st], [-plural], [+ argument] b. [1st], — [ plural], [- argument], [+ nominative] c. [1st], [- plural], [- argument], [- nominative]

FONGBE HAITIAN nyè mwen ùn mwen mì mwen

Consequently, in the incipient creole, there would be three homophonous forms for the first person singular pronominal lexical entries. The availability, in the incipient creole, of the lexical entries in (11) would enable the creators of Haitian who had both strong and weak pronominal forms in their original lexicons to reproduce these forms in the creole. However, using the same superstratum string to relabel several lexical entries copied from the substratum language(s) yielded redundancy in the newly created lexicon. Brousseau (1995a) thus further hypothesises that the three homophonous lexical entries in (11) were reduced to one, with their common features, yielding a single Haitian lexical entry unspecified for the features [a argument], where a is a variable that can take the values + or -, and [a nominative], where a is a variable that can take the values + or -. The reduced lexical entry is shown in (12) ( 12)

/mwen/: [ 1st], [- plural]

HAITIAN

The fact that the lexical entry in (12) is underspecified for the feature [a argument] also enabled the creators of Haitian, who had both strong and weak pronominal forms in their original lexicons, to produce these forms while speaking the creóle. Whether or not the first generation of Haitian native speakers was exposed to the data in (11) or (12), they nevertheless had no way of distinguishing between strong and weak forms on the basis of the data. Presumably, they observed the same form in all contexts where a pronominal was used by the adult population. Furthermore, Brousseau (1995a) points out

DO CREOLE LANGUAGES REALLY FORM A TYPOLOGICAL CLASS

219

that the context par excellence where the clitic and the strong forms were distinguished in terms of word order in the original grammar—that is, in nominalisations—had been abandoned in the early creole. It is thus reasonable to conclude that the first generation of Haitian native speakers could not deduce the availability of syntactic clitics based on the data that they were exposed to. Still according to Brousseau (1995a), presumably, the first generation of Haitian native speakers interpreted these data as in (13). (13)

/mwen/: [1st], [- plural], [+ argument]

HAITIAN

Thus, in modern Haitian, there are no syntactic clitics. The fact that syntactic clitics did not enter the creole can thus be derived from how relabelling is hypothesised to proceed in the case of functional category lexical entries in creole genesis. A second way by which a functional-category lexeme can acquire a label in creole genesis is through reanalysis. As will be seen in the next section, under specific circumstances, it may happen that such a lexeme cannot be relabelled during the relexification process. In this case, the copied lexical entry is assigned a null form. As has been proposed in Lefebvre and Lumsden (1992, 1994b), a functional category lexical entry that has been assigned a null form at relabelling may be signalled by a periphrastic expression. For example, a lexical entry having a temporal/aspectual meaning but a null phonological representation may be signalled by the use of an adverb with a similar meaning. The periphrastic expression may later become the phonological representation of the lexical entry initially assigned a null form, through the process of reanalysis. Such cases are reported in the literature (for an example from Tok Pisin, see Sankoff 1991). The fact that creoles are generally isolating languages can thus be explained through the relexification account of creole genesis described above. Since the functional-category lexemes of creole languages derive their phonological forms from major-category lexemes in the superstratum language, or from reanalysis, and since these categories are typically free morphemes, it is logical that creoles tend to be isolating languages (see Lefebvre and Lumsden 1994a, 1994b).5 5

Mufwene (1989: 124) accounts for the isolating character of creole languages by appealing to the notion of salience: "With regard to the issue made here, viz., explaining why periphrasis is generally preferred to inflections in PCs, I submit that salience should do." The proposal advocated in our research is somewhat similar to Mufwene's, for major categories may be

220

ISSUES IN THE STUDY OF PIDGIN AND CREOLE LANGUAGES

8 . 3 . Why do creole languages look simple? McWhorter (2001), states that creole languages are simpler than both their lexifier and susbtrate languages. On the relexification account of creole genesis assumed here, the issue of the alleged simplicity of creole languages can only be taken up in terms of a comparison of a creole with its substratum languages. This is thus the methodology that I will adopt in addressing the question addressed in this section. Are creole languages really simpler than their substratum languages? Or do they just happen to 'look' simpler? In the paragraphs that follow, I present a way of looking at the data that support the second alternative. In my view, creole languages only look simpler than their substratum languages. The fact that they look simpler than their substratum languages is that, due to constraints associated with the process of relexification, there are more covert categories in creole languages than there are in their substratum languages. By covert category, I mean a lexical entry that is required by universal grammar but that is phonologically null. In practical terms, this means that such a lexical entry has a syntactic function that can be argued for, but that it is not pronounced (see Lumsden 1995b). A case in point would be the accusative case in English. This case is required by universal grammar (see Chomsky 1981), but in English it is covert, unlike in other languages, such as Quechua, that have overt case morphology. Another relevant example involves the optional pronunciation of the complementiser that in English, as in John said 0 he would come. There is a consensus in the literature that, when the complementiser that is not pronounced in a sentence of the aforementioned type, the syntactic position is nonetheless filled by the features of this complementiser, and the covert complementiser plays a syntactic role in the structure of the clause. In light of this preliminary discussion, I now turn to the discussion of phonologically null lexical entries produced during the relexification process in creole genesis. It will be shown that phonologically null lexical entries result from constraints operating on relexification. According to Muysken (1981a: 62), relexification is semantically driven. "For relexification to occur, the semantic representations of source and target language entries must partially overlap; otherwise, the two entries would never be associated with each other. Other features of the two entries may, but need not, be associated with each other." In Lefebvre (1998a: 17), I take the position that, in relexification, copying may apply to all lexical entries and that it is viewed as "salient" when compared with minor categories. As is pointed out in Lefebvre (1998a: 48), however, the proposal in Lefebvre and Lumsden (1994a) is more specific, since it links the observed facts to the processes that generate them, namely, relexification and reanalysis, and to the linguistic material on which these processes apply in creole genesis.

DO CREOLE LANGUAGES REALLY FORM A TYPOLOGICAL CLASS

221

relabelling that is semantically driven. Thus, only those functional categories that have some semantic content (e.g. determiners, demonstrative terms, etc.) may be assigned a new label during relexification. Functional categories that have no semantic content (e.g. case markers, operators, etc.) are copied but they are not relabelled; they are phonologically null or covert. Practically speaking, this means that these lexical entries are not pronounced. As is pointed out in Lefebvre (1998a: 17-18), the claim that functional categories may be assigned a null form at relabelling is independently motivated by the fact that, in natural languages, functional categories required by universal grammar are not always spelled out, as we saw above on the basis of data drawn from English. In this respect, consider the Haitian and Fongbe nominal structures in (4), reproduced as (14) for convenience. (14)

krab [mwen 0] sa àson [nyè ton] éló crab me GEN DEM 'these/those crabs of mine'

a yo HAITIAN 6 lé FONGBE DEF PL (in question/that we know of)

The possessive phrase that follows the head noun of these structures comprises a pronoun and a case marker. The case marker is overt in Fongbe but covert in Haitian (see Lumsden 1991). Since case markers do not have semantic content, the Fongbe case marker could not be relabelled, and thus, the copied lexical entry from this substratum case marker was assigned a phonologically null form at relabelling. On the basis of syntactic tests, Brousseau and Lumsden (1992) argue that Fongbe ton has the properties of genitive case (= 's in English) rather than those of objective case (=of in English). Lumsden (1991) argues that the Haitian possessive phrase has the same properties as the corresponding Fongbe one. He thus identifies the phonologically null case marker as genitive. Another example of a functional category that could not be relabelled because it does not have semantic content involves the operator found in relative and f active clauses. This operator is lexical in Fongbe but it is covert in Haitian, as is illustrated in (15) involving factive clauses. (15)

Wá ɖé-è Jan wá o Jan rive a Rive 0 arrive OP John arrive DEF The fact that John arrived ... '

... ...

FONGBE HAITIAN (=(3) in Lefebvre 1994b)

222

ISSUES IN THE STUDY OF PIDGIN AND CREOLE LANGUAGES

The properties of the Fongbe operator are extensively discussed in Collins (1994) and in Kinyalolo (1993a). In Lefebvre (1998a: 203-205), it is argued that the Haitian null form in (15) has syntactic features that are manifested in the syntax of the construction, and that these features parallel those of Fongbe ɖé-è. The two sets of data presented above illustrate cases where a phonologically null lexical entry in the creole results from the fact that relabelling is semantically driven. There is another constraint involved in the process that may also yield phonologically null lexical entries in the creole. As was mentioned earlier, in Lefebvre and Lumsden (1994a), it is proposed that functional categories of the substratum languages that have some semantic content are relabelled on the basis of major category lexical items of the supertratum language that have some semantics in common and similar distributional properties. Relabelling is thus constrained by what the superstratum language has to offer in terms of appropriate phonetic strings to relabel a copied lexical entry. If no appropriate form is found, the copied lexical entry remains covert, that is without a label. This proposal accounts for differences observed between creoles formed from the same substratum languages but different superstrata. For example, French-based creoles of the Atlantic were able to reproduce the postnominal determiner of their substratum languages (see (4)) because French has an adverbial form that has the appropriate properties to relabel the copied lexical entry. Saramaccan, an English-based creole with the same substratum languages as Haitian (see Smith 1987), however, was not able to reproduce its substratum languages' postnominal determiner because, presumably, English does not have an appropriate form to relabel the copied lexical entry. On the other hand the lexical -self anaphor of the substratum Gbe languages was reproduced in the English- and Dutch-based creoles because, as is shown in (16), these superstratum languages have a -self anaphor. Since French does not have a -self anaphor, French-based creoles have a covert lexical entry in this case, as is illustrated in (17).6 (16)

6

Examples of creoles that have a reflexifive anaphor (pronoun + SELF) BERBICE DUTCH: -selfu (from Dutch -zelv) (Robertson 1993: 307) GULLAH: -self (from English -self) (Mufwene 1992: 169) SARAMACCAN -seéi (from English -self) (Veenstra 1996a: 43)

For an extensive discussion of these facts, see Lefebvre (1998a: 159-171). The idea that the lexical entry copied from the substratum -self anaphor could have been assigned a phonologically null representation at relabelling is attributable to John Lumsden (research seminar, Fall 1993). The further development of this idea is mine.

DO CREOLE LANGUAGES REALLY FORM A TYPOLOGICAL CLASS

223

Thus, phonologically null lexical entries in a creole may result from the fact that the superstratum language does not have an appropriate form to relabel a copied lexical entry [see chapter 9 for further discussion]. As can be seen from the above examples, there are more covert forms in the creole than there are in its substratum languages. As has been pointed out in Lumsden (1995b), this makes the creole lexicons look "simpler" than the original lexicons. Furthermore, lexical entries that are not required by universal grammar, and that cannot be relabelled due to either one of the two constraints discussed above, may simply be abandoned. This is the case, for example, of the logophoric pronoun of the Haitian substratum Gbe languages (see Lefebvre 1998a: 147). A logophoric pronoun is a pronoun that has no independent reference. Because they are not semantically independent, logophoric pronouns cannot be relabelled. There are no arguments that would support an analysis according to which there would be a null logophoric pronoun in modern Haitian. Therefore, in this case, it is simply assumed that the lexical entry has been lost. Cases of this type also make creole lexicons look simpler than the original lexicons. (For an extensive discussion on phonologically null forms in Haitian Creole, see Lefebvre 1998a: 378-381.) 8 . 4 . Apparent simplicity and hidden complexity In this section, I would like to call the reader's attention to some semantic interpretative facts showing that 'what you see is not always what you get' and that 'what you see is sometimes simpler than what you in fact get'. Consider the predicate cleft construction in (28). (18)

Wá w Jan wá. Se rive Jan rive. it.is ARRIVING it.is John arrive 'It is ARRIVING that John did.' (not e.g. leave)

FONGBE HAITIAN (=(4) in Lefebvre 1994b)

224

ISSUES IN THE STUDY OF PIDGIN AND CREOLE LANGUAGES

As is extensively shown in Lefeb vre (1990), in this construction, the clefted constituent may be assigned an intepretation that goes beyond what is actually found in the clefted phrase. For example, the clefted constituent in (19) may be assigned three different contrastive interpretations: one bearing on V, another one bearing on the VP, and a last one bearing on the internal argument of the verb. (19)

Se monje Jan monje pen an HAITIAN it.is eat John eat bread DEF 'It is EATING the bread that John did.' (not e.g. throw it away) 'It is EATING THE BREAD that John did.' (not e.g. wash the dishes) 'It is eating THE BREAD that John did.' (not e.g. eat the apple) (=(44) in Lefebvre 1990)

The example in (20) presents similar focal ambiguities. (20)

Se mache Jan mache al lekol HAITIAN it.is walk John walk to school 'It is WALK that John did to school.' (not e.g. run) 'It is WALK TO SCHOOL that John did.' (not e.g. run home) 'It is TO SCHOOL that John walked.' (not e.g. to the park) (=(23) in Larson and Lefebvre 1991: 251)

Finally, when the affected argument of a verb has been clefted, the contrastive interpretation of the cleft constituent bears either on the noun phrase, or on the whole VP. This is illustrated in (21). (21)

Se pen an Jan manje. HAITIAN it.is bread DEF John eat 'It is the BREAD that John ate.' (not e.g. the apple) 'It is EATING THE BREAD that John did.' (not e.g. wash the dishes) (=(53) in Lefebvre 1990)

The semantic interpretation facts in (18)—(21) are not directly accessible from the S-structures and they require semantic rules of interpretation that are not so 'simple'. They do far more than just establishing a one-to-one correspondence between the S-structures and their interpretations. Larson and Lefebvre (1991) analyse these facts in terms of quantification of events. These facts, and others of the same type that are discussed in Lefebvre (1998a, 2001c), show that some Haitian grammatical properties are more complex and certainly more opaque than a "simplicity" approach to creole languages would lead one to believe.

DO CREOLE LANGUAGES REALLY FORM A TYPOLOGICAL CLASS

225

8.5. McWhorter's list revisited McWhorter (2001) provides a list of fourteen features that he claims will never be found in a creole language. As he puts it: "Crucially: One would find a great many of the above features in the lexifier and substrate languages that were spoken by the creators of these creoles." In the theory of creole genesis advocated in the previous sections of this chapter, the creators of a creole do not have enough access to the superstratum language to learn the functional categories of that superstratum language. Thus, according to this approach, the lexifier language is not pertinent to explain the absence, in creole languages, of the list of items (almost all related to functional categories) provided by McWhorter. Only the substratum languages are pertinent to the discussion of this list. So, in the paragraphs that follow, I discuss McWhorter's list with respect to the substratum languages of Haitian, mainly Fongbe, and occasionally, other West African languages. Then, I propose a global evaluation of these features. None of the substratum languages of Haitian have ergative case; but even if they did, on the theory of relexification outlined in this chapter, ergative case would have been assigned a null form at relabelling, and thus, it would not be visible in the incipient creole. Gbe languages do have evidential markers (see Lefebvre and Brousseau 2002). As is shown in Lefebvre (1998a: 213-217), Haitian has a subset of them. Inalienably possessed objects must appear in the genitive case in Fongbe (see Lefebvre and Brousseau 2002). As we saw in (4), due to the semantic constraint on relabelling, genitive case is covert in Haitian, so there is no way to tell whether inalienably possessed objects occur in the genitive or in the objective case. Fongbe does not have switch reference, inverse or obviative marking. If it did, it is unlikely that these morphemes would have made their way into Haitian, because, provided that they have enough semantics to be relabelled, there may not be any appropriate French phonetic string to relabel the substratum morphemes. Fongbe, like the other Gbe languages, does not have verb raising to INFL, as we saw in Table 8.2, and thus, it does not manifest verb second phenomena, nor the syntactic asymmetries between matrix and subordinate clauses that go with them. As we saw in Table 8.2, verb raising to INFL (and eventually to a higher position in the syntactic tree) correlates with inflectional morphology of the verb. As we saw in Table 8.2, Haitian follows the pattern of its substratum Gbe languages with respect to this parametric option: neither have inflectional morphology of the verb. As for subjunctive marking, Gbe languages encode this mood by means of a preverbal marker. As we saw in

226

ISSUES IN THE STUDY OF PIDGIN AND CREOLE LANGUAGES

Table 8.1, this preverbal marker was reprodued in Haitian by relexification. Gbe languages present a few cases of syntactic clitic movement (see Lefebvre and Brousseau 2002). As was shown in section 8.3, syntactic clitics are not reproduced in a creole as a consequence of how relabelling proceeds in the case of functional-categoy lexemes. It follows that, unless a creole develops syntactic clitics, clitic movement will not be found in incipient creoles. McWhorter claims that creoles will manifest only an SVO word order. As is shown in Lefebvre and Brousseau (2002), Fongbe manifests a surface word order that is SVO in some contexts (mainly finite clauses) and OVS in others (mainly nominal and nominalised structures). In Lefebvre and Lumsden (1992), it has been proposed that word order in a creole will be established in the following way: Because the creators of the creole are aiming to reproduce the superstratum sequences they are exposed to and since they are able to identify the major category lexical entries, the word order of major category lexical items and major constituents in the creole will follow that of the lexifier language. However, because the creators of a creole do not have enough exposure to the superstratum language, they cannot identify its functional-category lexemes; when they relexify the functional-category lexemes of their native lexicons, they keep their original directionality properties. Hence, these items are predicted to have the same word order as in the substratum languages. The Haitian data presented in Lefebvre (1998a) show that this hypothesis is borne out. The data in Table 8.1 and in the nominal structures in (4) constitute examples in point. (For further discussion of this issue, see Lefebvre 1998a: 388-390.) Additional evidence for this claim comes from Berbice Dutch. Kouwenberg (1992) reports that Eastern Ijo, Berbice Dutch's main substratum language, is underlyingly an OV language. Dutch, the lexifier language, is also underlyingly an OV language. Berbice Dutch itself is a VO language. Kouwenberg explains this situation as follows: In Dutch simple clauses, the verb moves to INFL in such a way that the S-structure of Dutch simple sentences exhibit the order SVO. According to Kouwenberg, the creators of Berbice Dutch perceived this word order and hence established the word order SVO for the creole.7 Gbe languages used to have noun classes, still attested by frozen forms in the modern varieties. Whether these noun class prefixes were still productive at the time Haitian Creole was formed is unknown to me. The fact that Haitian Creole does not have noun class prefixes, however, suggests that noun class 7

As noted by Kouwenberg (1992), however, Berbice Dutch has postpositions. This should come as no surprise since Dutch also has postpositions. The fact that Saramaccan has postpositions (see Muysken 1987) when its English lexifier language does not, however, constitutes a counter-example to the proposal in Lefebvre and Lumsden (1992) on how word order is established in an incipient creole. The latter data require further investigation.

DO CREOLE LANGUAGES REALLY FORM A TYPOLOGICAL CLASS

227

prefixes were probably no longer productive in Gbe at the relevant time. This claim is supported by the fact that all the productive morphology of Gbe has been reproduced in the creole, as is extensively demonstrated in Lefebvre (1998a: 303-334, and the references cited therein) [see also chapter 10]. Finally, while Fongbe has phonological tones, Haitian Creole does not (see Cadely 1994). This terminates the discussion of the features in McWhorter's list as they relate to Haitian and its substratum languages. I now turn to a more global evaluation of the facts discussed above. I begin with the facts that are not in agreement with McWhorter's claim. The Haitian data involving the subjunctive and the evidential markers constitute counter-examples to McWhorter's claim. The fact that Haitian does not manifest verb second phenomena is irrelevant to its being a creole. As we saw earlier, the availability of this option in a particular grammar correlates with the availability of inflectional morphology in that particular grammar. Other languages, not identified as creoles, lack inflectional morphology, and hence, verb movement to INFL and, in some cases, to COMP. Chinese is a case in point. In turn, the presence of this feature in McWhorter's list is in contradiction with the property that he claims characterises the items in his list: "Crucially, none of these factors require inflectional morphology for their occurrence in a grammar, and thus their absence in not an epiphenomenon of isolating typology" (McWhorter 2001). Consequently, the features in this first group should be dropped from McWhorter's list of features that are excluded from creole languages. A second group of features comprises those that can be derived from the relexification account of creole genesis outlined in the previous sections. For example, all the features that are related to case marking (that is, lack of ergative case, lack of genitive case, and lack of particular case distinction for inalienable possession) are derivable from the semantic constraint on relabelling; from this constraint, case markers are not relabelled in relexification, and thus, they are predicted not to be overt in the incipient creole. Likewise, under the condition that they have enough semantics to be eligible for relabelling, markers or morphology involving switch, obviative or inverse reference cannot be relabeled for lack of appropriate material in the superstratum language. Similarly, the lack of clitic movement in creoles results directly from the lack of syntactic clitics in these languages. As we saw in section 8.3, the lack of syntactic clitics in a creole is derivable from the way relabelling is hypothesised to proceed in creole genesis. The fact that creoles are SVO languages is also derivable from the proposal concerning how word order is established in creole genesis contexts, even in

228

ISSUES IN THE STUDY OF PIDGIN AND CREOLE LANGUAGES

cases where contributing languages are SOV. So, all the features in this second group are derivable from a sound theory of how creole languages are created. Finally, the absence of tones in creole languages may be due to the mixed character of these languages. For example, as is argued in Brousseau (in preparation), the accentual system of Haitian represents a principled compromise between the tonal system of its Gbe substratum languages and the extremely simple accentual system of French. Likewise, the phonological system of Haitian represents a principled compromise between the phonology of its contributing languages. This is just like the relexified lexical entries which represent a principled compromise between the properties of the substratum lexical entries and those of the superstratum language. On the basis of this global evaluation of McWhorter's list, my conclusion is the following: The first group of features should be removed from the list in light of the Haitian counter-examples that have been presented. The second and third groups of features should be retained. These are the ones that can be derived from the relexification account of creole genesis presented in earlier sections of this chapter. 8.6. Conclusion The central thesis advocated in this chapter is that creole languages do not form a typological class. It was shown that the creators of creole languages are adult native speakers who use the properties of their own lexicons and grammars in creating the creole. The bulk of a Creole's lexicon is thus created through the process of relexification. This account of creole genesis predicts that creoles reproduce the properties of their substratum languages, in such a way that creoles from different geographical areas will manifest the same type of differences among themselves as their respective substratum languages do. It was proposed that what appears to unite creole languages from all geographical areas is the main process—relexification—by which they come about. In spite of this strong conclusion, it was shown that creole languages appear to share at least one feature: they tend to be isolating languages. It was argued that this property of creole languages results from the relexification account of creole genesis. Regarding the issue of simplicity per se, it was shown that, due to constraints associated with the process of relexification—the fact that relabelling is generally semantically constrained, and the fact that relabelling is, in particular, constrained by what the superstratum has to offer to relabel a copied lexical entry—there are more covert lexical entries in creole languages than there are in their substratum languages. This makes creoles "look simpler" than the original lexicons. Semantic interpretation data were presented showing that apparent simplicity may hide effective complexity. Finally, the evaluation of the features proposed by McWhorter (2001) as being absent from creole languages led to an interesting

DO CREOLE LANGUAGES REALLY FORM A TYPOLOGICAL CLASS

229

conclusion. Putting aside the few features that should be removed from the list, it is possible to derive the others from a theory of creole genesis based on the major process that is at work in creole formation. A list that comprises features that seem heterogeneous at first glance thus acquires some justification when related to the process that creates the languages under discussion.

CHAPTER 9

THE INTERPLAY OF RELEXIFICATION AND LEVELLING IN CREOLE GENESIS AND DEVELOPMENT* Dialect levelling, as discussed in the literature on dialects in contact (see e.g. Domingue 1980, 1981; Siegel 1995, 1997; Trudgill 1986; etc.), refers to the reduction of variation between dialects of the same language in situations where these dialects are brought together. As Siegel (1997: 128) puts it, "dialect differences are reduced as speakers acquire features from other varieties as well as avoid features from their own variety that are somehow different. This may occur over several generations until a stable compromise dialect develops". Dialect levelling is thus a social process that can be described as a negotiation between speakers of different dialects of a given language. The process results in agreement on a form or on the properties of a given form in the context where either competing forms or competing properties for a given form coexist. Examples of well-documented cases of dialect levelling in non-creole communities include Bhojpuri as spoken in Mauritius (Domingue 1980, 1981) and English as spoken in America, Australia, New Zealand and South Africa (Trudgill 1986). In the literature on pidgins and creoles, it has been proposed that the process of levelling plays a role in the development of these languages (see e.g. Harris 1991; Mufwene 1990, 1994c, 1996b; Mühlhäusler 1980; Siegel 1997). For example, Siegel (1997: 131-132) asserts that: Mixing and levelling may [...] be important in the development of stable pidgin and creole languages [...]. When the various versions of the superstrate are [...] used as the main means of communication among speakers of different substrate languages (in other words, when vernacularisation occurs), and when these speakers begin to view themselves somehow as a 'community', then levelling begins. The content of this chapter was originally published in Linguistics (Lefebvre 2001a) and presented at the Linguistic Society of America (Society for Pidgin and Creole Linguistics) meeting in January 1999, in Los Angeles, under the title 'Dialect levelling in creole communities'. I would like to thank Renée Lambert, Catherine Léger and Olivier Tardif for assisting me in doing some of the research that led to this chapter. Thanks to Andrée Bélanger for formatting the manuscript and to Catherine Léger for proofreading it. The research in this chapter was financed by SSHRCC and FCAR. I would like to thank Bernard Comrie and Jeff Siegel for their very useful comments on an earlier draft, as well as a number of (anonymous) scholars for their questions and comments. I am also grateful to Christine Jourdan and Jeff Siegel for fruitful discussions on several of the issues raised here.

232

ISSUES IN THE STUDY OF PIDGIN AND CREOLE LANGUAGES

The above references all acknowledge the fact that speakers of incipient creoles manifest variation among themselves. This raises two questions. The first one can be formulated as follows: What is the source of the variation found in incipient creoles? There is a consensus in the literature to the effect that communities where pidgin and creole languages emerge generally involve several substratum languages, spoken by the majority of the population, and one superstratum or lexifier language, spoken by a relatively small social group (e.g. Whinnom 1971). One option is to say that the variation observed in an incipient creole comes from dialectal variation found in the superstratum language. Another option is to say that it comes from the variation manifested between the various contributing substratum languages. Building on Lefebvre (1998a), I argue below that the bulk of the variation found in an incipient creole is attributable to the variation manifested between the substratum languages (sections 9.3 and 9.4), and even to the variation found within the individual substratum languages of this incipient creole (section 9.5). The second question can be formulated as follows: How does the variation observed between the various substratum languages of an incipient creole get transferred into the creole? I demonstrate below that this is achieved through the process of relexification. This chapter addresses the topic of levelling in creole communities as it relates to the relexification account of creole genesis developed in the course of the UQAM Haitian projects (see Lefebvre 1998a, and the references cited therein). I would like to acknowledge the important contribution of John Lumsden in the articulation of these two processes, that is relexification and levelling (see mainly Lefebvre and Lumsden 1989, 1994a, 1994b; Lumsden and Lefebvre 1994). In this chapter, I am building on theoretical work that we did together. I take full responsability for the implementation of the data within this framework. I will assume the general theory of creole genesis extensively outlined and documented in Lefebvre (1998a) on the basis of Haitian Creole and its contributing languages. In order to preserve the autonomy of this chapter, however, I will summarise the pertinent facts and assumptions in each section. The question of how relexification and levelling are implemented in everyday life is not discussed here (for a discussion of this topic, see Durie in preparation). The processes identified above as playing a role in the genesis and development of creole languages take place in communities and hence, they interact with the social components which define the features of these communities. Although social factors are not discussed here, I assume that they interact with linguistic processes (particularly levelling) in a way similar to that described in Labov's (and his associates') meticulous work on the interplay of linguistic and social

THE INTERPLAY OF RELEXIFICATION AND LEVELLING

233

factors in communities. Principles that guide levelling are not discussed either (for preliminary discussions of this point, see Jourdan 1985 and Siegel 1997). The chapter is organised in the following way: Section 9.1 summarises the theoretical framework referred to above. The basic idea is that, while relexification of various substratum lexicons on the basis of a single superstratum language provides the speakers of an early creole community with a common vocabulary, it is the major cause of variation in an incipient creole. In this view, levelling is hypothesised to operate on the output of relexification, thus reducing the variation produced by the relexification of several lexicons. Section 9.2 summarises the situation that prevailed in Haiti when Haitian Creole was formed. Sections 9.3, 9.4 and 9.5 implement the theory outlined in section 9.1 on the basis of a subset of data drawn from Haitian Creole and its contributing languages. A full discussion of the general research methodology may be found in Lefebvre (1998a: 52-78). In this chapter, I consider cases involving distinctive properties of specific lexical entries in the substratum languages and compare them with those of the corresponding lexical entries in the creole. Third person plural pronouns and plural markers, ways of encoding reflexivity and deictic terms will be discussed in turn. The discussion of these data goes beyond that in Lefebvre (1998a) and it considers questions and comments raised since then by various scholars. Section 9.6 concludes the chapter. 9 . 1 . Relexification and levelling This section shows how the process of relexification is predicted to account for the variation found in the early creole communities and how levelling is hypothesised to operate on the variation produced by relexification of various substratum languages on the basis of a single lexifier language. As we saw in section 3.1.1, a relexified lexical entry thus has the semantic and syntactic properties of that of the lexicon it has been copied from and a phonological representation derived from a phonetic string in the lexifier language.1 In line with the terminology adopted in Lefebvre and Lumsden (1994a, 1994b), I will refer to the process of assigning a new phonological representation to a copied lexical entry as relabelling. Relabelling proceeds from phonetic strings found in the superstratum language. Furthermore, it is hypothesised that, if no form is available in the lexifier language to relabel a

1 The question of how phonetic strings of the superstratum language are phonologically interpreted by the substratum speakers is discussed at length in Brousseau (in preparation) on the basis of Haitian Creole.

234

ISSUES IN THE STUDY OF PIDGIN AND CREOLE LANGUAGES

copied lexical entry, this lexical entry may be assigned a null form at relabelling.2 According to Muysken's (1981a: 62) proposal, relexification is semantically driven: "For relexification to occur, the semantic representations of source and target language entries must partially overlap; otherwise, the two entries would never be associated with each other. Other features of the two entries may, but need not, be associated with each other." Muysken's idea of partial semantic overlap between the source and target lexical entries is preserved by specifying that the meaning of the phonetic string selected to relabel a copied lexical entry is deduced from its use in specific semantic and pragmatic contexts. In Lefebvre (1998a), it is assumed that copying applies to all lexical entries and that relabelling is semantically driven. Thus, only functional categories with some semantic content (e.g. determiners, demonstrative terms, etc.) may be assigned a new label when relexified. Functional categories which have no semantic content (e.g. case markers, operators, etc.) can be copied but not relabelled. They are assigned a null form at relabelling. As is noted in Lefebvre and Lumsden (1994b), practically speaking, this means that these lexical entries are not pronounced.3 In Lefebvre and Lumsden (1994a), it is proposed that the functional category lexical entries of the copied lexicon are relabelled using major category lexical items of the superstratum language and that the distributional properties of the superstratum form must be similar to those of the copied lexical entry. Relabelling of functional category lexical entries is thus constrained by what the superstratum language has to offer in terms of major category lexical items. The semantics and distribution of a lexical item of the superstratum language have to be appropriate in order to provide a phonetic matrix to a copied functional category lexical entry. The constraint imposed by the superstratum language on the relabelling of functional category lexical entries is discussed in Lefebvre (1998a). Constituent order in the superstratum language also acts as a constraint in establishing constituent order in the incipient creole. Such a case is documented in detail in Lumsden (1991) (see also section 9.5 of this chapter). In Lefebvre and Lumsden (1989, 1994a, 1994b) (see also Lefebvre 1998a: 10, 30-41), it is claimed that, in creole genesis, the process of relexification is used by speakers of the substratum languages as a tool for acquiring a second language, the superstratum language. As is pointed out in Lefebvre and Lumsden (1989), this account is a further development of the 2 For other representations of the process, see Mous (1995, in press) and Muysken (1981a). For a general discussion on the formal representation of relexification, see Lefebvre (1998a: 15-19, 22, 27, 384-386). 3 For an extensive discussion on null forms in creoles, see Lefebvre (1998a: 17-18, 37-38, 44-45, 78, 108-110, 128, 164, 375-386).

THE INTERPLAY OF RELEXIFICATION AND LEVELLING

235

second language acquisition theory of creole genesis. For example, Alleyne (1971, 1980), Andersen (1980), Chaudenson (1993), Mufwene (1990), Schumann (1978), Thomason and Kaufman (1991) and Valdman (1980) have proposed that pidgin and creole languages constitute a crystallised and incomplete stage of second language acquisition. As is discussed in Lefebvre (1998a), without relexification, this approach to creole genesis does not explain why creole languages have crystallised the way they have (see also Lefebvre 1984; Lefebvre and Lumsden 1989). The relexification hypothesis does explain why creole lexicons reflect the properties of both their superstratum and substratum source languages the way they do. This raises the question of whether relexification plays a role in second language acquisition. As is discussed in Lefebvre (1998a: 33-35, and the literature cited therein), substratal features in radical creoles have long been considered as cases of transfer, shown to play a major role in second language acquisition (see e.g. Weinreich 1953). On the one hand, in the theory of creole genesis advocated here, the type of data claimed to be associated with the notion of transfer in creole genesis corresponds to the result of the process of relexification (see Lefebvre 1998a). On the other hand, Lumsden (1999a), Mous (1995) and Siegel (1997) argue that, in some cases of second language acquisition, relexification (or paralexification, using Mous' term) plays a role. Thus, in both creole formation and more simple cases of second language acquisition, relexification may be argued to be used as a means of acquiring another language. As is discussed in Lefebvre (1998a, and the references cited therein), the amount of relexification in both cases appears to be a function of the amount of exposure to the lexifier language. Little exposure to L2 triggers recourse to relexification, whereas more exposure to L2 increases the degree of successful acquisition.4 As has been pointed out many times (e.g. Lefebvre and Lumsden 1989, 1994a, 1994b), relexification is a mental process and thus an individual process. Typically, situations where pidgins and creoles emerge involve several substratum languages, as we saw above. Each individual of the substratum languages relexifies his or her own lexicon in the early creole community. Relexification of several lexicons on the basis of a single superstratum language provides the early creole community with a common vocabulary. But, as is also pointed out in Lumsden and Lefebvre (1994), in relexification, speakers of 4

It has been argued many times that the process of relexification plays a major role in the formation of mixed languages (see e.g. Bakker 1992; Mous 1994, in press; Muysken 1981a). It has also been argued by the same authors that second language acquisition is not involved in the formation of mixed languages, for they are created by bilingual speakers. As is pointed out in Lefebvre (1998a: 29), relexification may sometimes be used as a tool in second language acquisition (but not in every cases). Being a cognitive process, relexification is autonomous and it may be used in different situations of language creation, whether acquisition of L2 is involved (as in creole formation) or not (as in mixed language formation).

236

ISSUES IN THE STUDY OF PIDGIN AND CREOLE LANGUAGES

various substratum languages reproduce the idiosyncratic semantic and syntactic properties of their own lexicons; thus the product of relexification is not uniform across the early creole community. In this view, the variation found in an early creole community is the product of relexification of various substratum lexicons. For example, relexification of the lexicons of languages X, Y and Z on the basis of a single superstratum language will yield three slightly different lexicons in an incipient creole. This is schematically represented in (1).

The features that are common to all the relexified lexicons (that is, to all the substratum languages) will most probably be maintained in the creole (see also e.g. Gambhir 1988: 77; Harris 1991: 199; Sankoff 1984; Siegel 1997; Singler 1988; Thomason and Kaufman 1991; Trudgill 1986: 98). The idiosyncratic features, however, are those that are subject to levelling. The criteria governing selection among competing forms require further research (see Mufwene 1986, 1990). A preliminary discussion of this point may be found in Siegel (1997). In the scenario of creole genesis presented in Lumsden and Lefebvre (1994), relexification occurs when the speakers of the substratum languages are targeting the superstratum language. It is hypothesised that, when they stop targeting the lexifier language and start speaking to each other using the relexified lexicons, that is, the early creole, they begin learning aspects of the incipient creole lexicons from each other and they begin levelling out the differences between the relexified lexicons. This is conceptually in line with the authors cited in the introduction to this chapter. The originality of the above proposal lies in the claim that, in this case, levelling reduces the variation produced by the relexification of the various substratum lexicons. It is thus hypothesised that, in the context of creole genesis, levelling operates on the output of relexification. As is mentioned above, the process of levelling starts when the speakers of the incipient creole (thus the creators of the creole) begin using the creole to communicate among themselves. By hypothesis, the creators of the incipient creole themselves may get involved in levelling. This does not mean, however, that the process of levelling takes place only in the initial formative stage of the creole nor that it levels out within a few years all the variation in the incipient creole. As will be seen in the data sections (sections 9.3 to 9.5), even if there are early cases of levelling, there are also cases which take much longer to be resolved. Furthermore, as is illustrated in Lefebvre (1998a), there are cases of

THE INTERPLAY OF RELEXIFICATION AND LEVELLING

237

variation in modern Haitian that parallel variation found in one of its substratum languages, which shows that, even after approximately three hundred years, levelling has not affected all parts of the lexicon of this language. The reason why some differences are levelled out early, others later, and still others not at all in the development of a creole remains a topic for future research. However, it should be pointed out that there is no reason to believe that the situation observed in creole communities with respect to variation is any different from cases of linguistic variation observed in non-creole communities (see e.g. Labov 1972a). Likewise, there is no reason to believe that the process of levelling, a case of linguistic change, should apply differently in creole communities than in noncreole communities. We know from the literature on linguistic change that some changes may occur rather rapidly, while others take more time to be completed. In the view advocated above, while the purpose of relexification in creole genesis is to provide speakers of different languages with a common vocabulary, the purpose of levelling is to reduce variation between the various dialects (or even idiolects) produced by relexification of different substratum lexicons. Finally, as has been pointed out in Lumsden and Lefebvre (1994: 8), "Since the process of reconciling the various dialects of the creole created by relexification of the various substratum languages must depend on the interactions of these dialects in the community, it follows that different communities will arrive at a somewhat different compromise". Evidence that this is the case may be found in regional dialect differences observed in a given creole (for Haitian, see Orjala 1970; Romain 1959; Tinelli 1970) or in differences among creoles formed on the basis of similar languages (e.g. both Haitian and Martinican creoles arose in historical situations involving basically the same languages (see Singler 1993b, 1996); nevertheless, these two languages present lexical differences (see e.g. Goodman 1964), which shows that levelling did not produce exactly the same results in these two communities). The above proposal provides a sound explanation for what has been referred to in the literature on creole studies as the "cafeteria principle"—a term first used by Dillard (1970) which was borrowed by Bickerton (1981: 49). As Bickerton (1981: 49) puts it, "As things stand, we are asked to believe that different African languages contributed different rules and features to particular creoles [...]. It is [...] absurd to suppose that a creole could mix fragments of Yoruba, Akan, Igbo, Mandinka, and Wolof [...]". On the one hand, the proposal that relexification plays a major role in the formation of creole languages accounts for the fact that specific features of the various substratum languages will be manifested in the early creole. On the other hand, the proposal that levelling operates on the output of relexification provides a principled explanation for the observation that different substratum languages may

23 8

ISSUES IN THE STUDY OF PIDGIN AND CREOLE LANGUAGES

contribute features to a given creole. This is because, as is pointed out in Lefebvre (1998a: 394), in cases of competing forms, it is not always the same linguistic group that wins. For example, Lefebvre (1998a: 219-246) shows that, in the Haitian case, Fongbe speakers were quite successful with regard to the clausal determiner, whereas the Ewe speakers won in other cases (see also section 9.3 of this chapter). Finally, as is pointed out in Lumsden and Lefebvre (1994), the role of levelling in creole genesis accounts for the fact that the properties of some specific lexical entries of the creole may depart from those of the corresponding lexical entries in the individual substratum languages. These points will be illustrated in the data sections (sections 9.3 to 9.5). 9.2.

The linguistic situation in Haiti at the time Haitian Creole was formed On the basis of the historical study that he carried out on the economy, demography and linguistic diversity of early Haiti, Singler (1996) provides evidence that Haitian Creole was probably created between 1680 and 1740. He demonstrates that the bulk of the Haitian population, at the time Haitian Creole was formed, was adult. Based on this fact, Singler (1996: 199) concludes that the principal agents of creole genesis must have been adults. As is documented in Singler (1996, Table 3), the majority of the Haitian population during that period was of African origin and people of mixed race; thus day-to-day contacts between French speakers and the Haitian-African population were very limited. Furthermore, Singler (1993b, Tables 3 and 4) shows that the Africans were from the following language families: Atlantic, Mande, Kwa (including Akan and Gbe), Gur, Nigerian Benue-Congo, Ijoid and Bantu. These African languages, although not entirely disparate from a typological point of view, are mutually unintelligible (see Lefebvre 1998a: 58-61, and the references therein). Thus, as is pointed out in Lefebvre and Lumsden (1994b), the situation that prevailed in Haiti between 1680 and 1740 presented the two basic prerequisites for the emergence of a creole language. First, the community was multilingual, speaking languages that were not mutually intelligible. Given the number of languages present, it was obviously in need of a lingua franca. This lingua franca was needed to permit communication not only between the speakers of the substratum languages and those of the superstratum language, but also between the speakers of the substratum languages themselves (see e.g. Foley 1988; Lefebvre 1998a: 1; Singler 1988:47; Thomason and Kaufman 1991). Second, the African people could not have had much direct access to native speakers of French during this period, given the figures provided by Singler. Thus, between 1680 and 1740, the social conditions in Haiti were such as to permit the development of a creole language. In the view advocated in Lefebvre and Lumsden (1994a, 1994b), the first step towards the creation of the

THE INTERPLAY OF RELEXIFICATION AND LEVELLING

239

lingua franca was achieved by relexifying several African lexicons5 using a single language, in this case, French. This brings me to the question of what type of French data the creators of Haitian were exposed to. This topic is documented at length in Lefebvre (1998a: 62-65), based on the literature. Here, I only summarise the most salient points. What variety of French did the French colonists speak in Haiti? There is a consensus in the literature (see e.g. Faine 1937; Hull 1979; Valdman 1978; Valdman (ed.) 1979; etc.) that the bulk of the French colonists who went to the Caribbean in the 17th and 18th centuries were from Western and Central France (i.e. Normandy, Île-de-France (Paris), Poitou, Saintonge, Brittany). The dialects spoken in this area of France constitute the langue d'oïl (as opposed to the langue d'oc). These dialects were not necessarily mutually intelligible among themselves or with the variety of French spoken in Île-de-France (Paris). (For various views on this matter, see the papers in Mougeon and Beniak (eds) 1994.) But, as noted by Meyer-Lübke (1909: 123), these regional dialects were influenced by Parisian French: "On sait très bien que, dans les provinces de I'Ouest, les dialectes ont subi, plus profondément que partout ailleurs en France, et même avant l'époque de la colonisation de l'Amérique, l'influence de la langue officielle." [It is very well known that, in the Western provinces, more than anywhere else in France, the dialects were influenced by the official language, even before the colonisation of America.] Furthermore, there appears to be a consensus in the literature on the characteristics of the French colonists of the 17th and 18th centuries (see e.g. Chaudenson 1992; Mougeon and Beniak (eds) 1994; Poirier 1994; Rivard 1914; Wolf 1994; etc.). First, the colonists were mainly from the domaine d'oïl. Second, even though they were born in the provinces, they had lived in cities before leaving France. Third, most of them spoke French. Finally, about half of them were literate. According to Juneau and Poirier (1973: 191-193), the variety of French spoken by the colonists was close to the variety of French spoken in Île-de-France in the 17th and 18th centuries: "Les traits syntaxiques sont facilement rattachables à des tendances de la syntaxe de l'époque. Il en est de même pour les traits morphologiques." [The syntactic features can very easily be linked to syntactic tendencies of that time. The same is true of the morphological features.] Wolf (1991) also draws a similar conclusion.6 There is also a consensus in the literature that the French colonists who went to North America in the same period spoke the same languages as those who went to the Caribbean (see Juneau 1972; Valdman 1978; Valdman (ed.) 1979; etc.). Various authors hypothesise that the 17th- or 18th5 Fongbe, a Gbe language of the Kwa family, was chosen as a case study for reasons extensively discussed in Lefebvre (1998a: 52-77). 6 For extensive discussions of these issues, see Mougeon and Beniak (eds) (1994) and the papers and references therein.

240

ISSUES IN THE STUDY OF PIDGIN AND CREOLE LANGUAGES

century French spoken by the French colonists, regardless of whether they went to North America, the Caribbean or Africa, was similar. For these reasons, in Lefebvre (1998a: 62), it is assumed that the French colonists in Haiti spoke a variety of French similar to that referred to as 17th- or 18th-century French in dictionaries and grammars. It is perhaps worth noting that, as is pointed out in Lefebvre (1998a: 63), some scholars who have worked on French-based creoles have adopted a different methodology which consists in looking for traces of French regional dialects in the creoles. For example, Picard, the French dialect spoken in Picardy, is often cited as the source of a creole form or feature. According to Juneau (1972), however, only 2% of the French colonists sent to the Americas were from Picardy. Consequently, it is unlikely that the bulk of the African population in the French Caribbean was exposed to this dialect of French in a way that would have had a significant impact on the French-based creoles. Furthermore, it is likely that, in the colonies, the French colonists who spoke regional dialects natively used the variety of 17th- or 18th-century French that they had in common, as attested in the historical work of the authors cited above. Given these considerations, I assume that the creators of Haitian were mainly exposed to the variety of French spoken in Île-de-France in the 17th and 18th centuries. The creators of Haitian, who spoke the various mutually unintelligible African languages enumerated above, were thus exposed to a variety of 17th- or 18th-century French, which probably contained regionalisms from the dialects of the langue d'oil In line with the general theory of relexification outlined in section 9.1, I take this variety of French to have provided the phonetic matrices for the relabelling of the lexical entries of the African lexicons and to have imposed the surface constraints upon this process (as discussed in section 9.1). The phonological representations of some Haitian lexical entries directly reflect this situation. For example, the Haitian verb bay 'to give' is derived from the French form bailler 'to give', no longer in use in modern standard French. Another example in point is provided by Lumsden (1991), who shows that the creators of Haitian Creole abandoned the prenommai complement of their original grammar in order to conform to the constituent order of French, where the nominal complement always follows the nominal head (e.g. la maison de Pierre 'the house of Peter'). So, the variety of French the creators of Haitian Creole were exposed to provided the superstratum material on which Haitian Creole was formed. Lefebvre's (1998a) extensive comparison of the lexical properties of Haitian Creole with those of its contributing languages argues, however, that the bulk of the variation found in modern Haitian Creole does not reflect the

THE INTERPLAY OF RELEXIFICATION AND LEVELLING

241

variation found in the variety of French that the creators of Haitian were exposed to, but rather to the variation found in the substratum languages. I refer the reader to Lefebvre (1998a, and the references cited therein) for a thorough demonstration of this point across a significant portion of the lexicon. The examples discussed in sections 9.3 to 9.5 of this chapter illustrate this point. This situation is congruent with the fact that, as was mentioned earlier, the creators of Haitian did not have enough exposure to French to acquire the properties of its lexical entries or its syntax. In light of this historical background, I now turn to the discussion of the data. 9 . 3 . Third person plural pronouns and plural markers This section discusses the variation between the substratum languages of Haitian with respect to the ways third person plural pronouns and plural markers are encoded. I look at how the lexicon of Haitian Creole compares with the lexicons of its substratum languages. The data show that, in this case, levelling must have taken place early in the development of Haitian. The content of this section builds on Lefebvre (1998a: 84-87), based on preliminary work in Lumsden and Lefebvre (1994). Fongbe has two distinct morphemes to encode the third person plural pronoun and the plural in noun phrases. This is shown in (2). (2)

a.

àson crab PL 'the crabs'

lέ

b.

Yé yì. 3pl leave They left.'

FONGBE

In contrast, Ewe has the morpheme wo, encoding both notions (see e.g. Wallace 1995: 10; Westerman 1930: 47, 58). This situation is also observed in other West African languages: Yoruba (see Goodman 1964:45-46; Migeod 1911: 170), Mandingo (see Delafosse 1929: 175), Igara and Igbira, and Vai, a Mande language (see Mufwene 1986: 138). Haitian differs from Fongbe in this case and follows the pattern manifested by Ewe and the other West African languages mentioned above. As is shown in (3), the Haitian morpheme yo serves both as a plural marker and a third person plural pronoun. (3)

a.

krab yo crab PL 'the crabs'

b.

Yo patí. 3pl leave 'They left.'

HAITIAN

As Mufwene (1986) asks, how can the difference between Haitian and Fongbe be accounted for within the framework of the relexification hypothesis? As is suggested in Lefebvre (1998a) and Lumsden and Lefebvre (1994), the answer to

242

ISSUES IN THE STUDY OF PIDGIN AND CREOLE LANGUAGES

this question lies in the interaction between relexification and levelling as it is hypothesised to apply in creole communities (see section 9.1). With Goodman (1964) and Sylvain (1936), let us hypothesise that the phonological representation of Haitian yo is derived from the strong form of the third person plural pronoun eux in French. As is argued in Lefebvre (1998a: 84-87), this is a likely hypothesis. First, Fongbe yé, Ewe wo, etc. share some element of meaning with French eux: all serve as a third person plural pronoun. Second, according to Brousseau (in preparation), the phonological derivation of French eux as Haitian yo is a predictable one. Note that, as is pointed out in Lefebvre (1998a: 411, note 4), although the French form eux is pronounced [yo] in Gascon and Auvergne French (see Goodman 1964), these French dialects were not represented in Haiti at the time Haitian Creole was formed (see section 9.2). Thus, it is unlikely that Haitian yo was phonologically derived from [yo] in Gascon or Auvergne French. Third, as is shown in Lefebvre (1994a, 1998a: 86), the substratum pronominal forms and French eux share distributional properties. The pronoun eux in French occurs as an emphatic form before a clitic, as in (4). (4)

Eux, ils mangent du riz chaque jour. them, they eat PARTrice every day Them, they eat rice every day.'

FRENCH (=(40) in Lefebvre 1994a)

Furthermore, in French, the third person plural pronoun eux also occurs as an emphatic pronoun after a nominal constituent, as is shown in (5a). Note the neutralisation in gender that extends to ils as well as to eux in (5b), also discussed in Gougenheim (1973). (5)

a. b.

Les gars, eux, ils... DEF.PL guy them they 'The guys, them, they...' Les filles, eux, ils... DEF.PL girl them they 'The girls, them, they... '

FRENCH POPULAR FRENCH (=(41) in Lefebvre 1994a)

In short, the French form eux was arguably a suitable candidate for relabelling the third person plural pronoun of Fongbe, Ewe and other related languages, but as is also noted in Lefebvre (1998a: 86), the Haitian form yo does not derive all of its properties from French eux. For example, eux cannot serve as a plural marker in French nominal structures. With these assumptions in mind, Lumsden and Lefebvre (1994) proposed the following general account of the historical derivation of Haitian yo. On the one hand, the Ewe speakers and speakers of similar languages had one

THE INTERPLAY OF RELEXIFICATION AND LEVELLING

243

lexical entry to relexify. Relexification of this lexical entry yielded yo, serving both as a third person plural pronoun and a plural marker in the early Haitian lexicon. On the other hand, the Fongbe speakers had two forms to relexify. Presumably, Fongbe speakers also relexified the third person plural pronoun of their lexicon as yo. As a result of interacting with other speakers of early Haitian, who had relexified their lexicons on the basis of Ewe-type lexicons, Fongbe speakers noticed that yo was also used to encode plural in nominal structures. As a result of levelling, they abandoned the form lέ of their own lexicon and extended the usage of yo so as to serve also as a plural marker in noun phrases. Presumably, this is an early case of levelling for, to the best of my knowledge, there is no attested dialect of modern Haitian that has two distinct forms for the third person plural pronoun and the plural marker. An alternative proposal was submitted to me for discussion. According to this proposal, the Fongbe speakers could have relexified both the third person plural pronoun and the plural marker occurring in noun phrases as yo on the basis of the two surface positions of French eux (see (4) and (5)), thus yielding two homophonous lexical entries in the early Haitian lexicon. This proposal is a likely one for, as is shown in Lefebvre (1998a), there are cases where different substratum lexical entries can be argued to have been relexified from the same French form. This proposal is also an attractive one for, if confirmed, it would illustrate a case where a difference between the substratum lexicons would have been eliminated from the creole as a result of the restricted number of available forms in the lexifier language. The plausibility of this scenario is weakened, however, in view of the following points. First, as we saw in section 9.1, relabelling is semantically driven; thus functional categories that have some semantics may be relabelled (e.g. the definite determiner), while those that do not have semantic content (e.g. case markers) may not be relabelled. Do plural markers in noun phrases have enough semantics to undergo relabelling? Having thought about this question for a long time, I still have no definite answer, although I tend to think that they do not. But, for argument's sake, let us suppose that, since plural markers have some semantics by virtue of being plural, they can be relabelled. Then, the question we should ask is whether or not the third person plural pronoun eux occurring in (5) is an appropriate form to relabel the Fongbe plural marker lέ. French eux in (5) and Fongbe lέ share the property of being plural and of occurring postnominally. However, while eux is pronominal, hence referential, lέ is not: it is only plural. While I can easily conceive of a third person plural pronoun being relabelled on the basis of a third person plural pronominal form (as in the first scenario), I can hardly conceive of a plural marker being relabelled on the basis of a third person plural pronominal form (as in the second scenario). Pending an independent theory on lexifier

244

ISSUES IN THE STUDY OF PIDGIN AND CREOLE LANGUAGES

language processing in second language acquisition, it is not useful to pursue the discussion of this point any further here. Given the problems raised by the second scenario, however, I conclude for the time being that the first scenario is the most plausible one. A third scenario can yet be considered, according to which levelling would have taken place between speakers of different African languages prior to relexification. On this view, at the time of relabelling, only one lexical entry having both functions for, say, Ewe and Fongbe speakers, had to be relabelled. Although this proposal is attractive, I think it is unlikely. Recall from the introduction to this chapter that levelling is a process that takes place when varieties of the "same language" are brought together. These varieties have to be mutually intelligible and they have to be the means of communication between the speakers for levelling to take place. As we saw in section 9.2, the various West African languages that were brought together in early Haiti, although not completely disparate from a typological point of view, were not mutually intelligible. Moreover, as is argued in Capo (1984), the Gbe languages constitute a dialect cluster or continuum, the two extremes of which (Ewe and Fongbe) are not mutually intelligible. Although one cannot exclude a priori the possibility that speakers of more similar Gbe dialects (e.g. Fongbe and Gungbe or Fongbe and Ajagbe) tried to communicate among themselves in their native tongues in spite of their differences, the hypothesis that the West African speakers, in early Haiti, learned each others languages in order to communicate among themselves is not supported by the historical facts. As we saw in section 9.2, the languages involved were numerous and their speakers needed a lingua franca (that would provide them with a common vocabulary) to help them communicate among themselves. This is why Haitian Creole was created. On this view, and as is discussed at length in Lefebvre (1998a), the African population in early Haiti was busy creating this lingua franca. Their target was French rather than the other West African languages present. Thus, on the basis of known historical facts, if any levelling took place prior to relexification, its impact on the creole was minor, if at all noticeable. Pending further theoretical work on language processing in second language acquisition and given what we know on the historical situation in Haiti, I conclude that the first scenario is the most plausible one; that is, the interaction between relexification and levelling, as hypothesised in the first scenario above, accounts for the fact that, in this particular case, Haitian differs from Fongbe and pairs with languages of the Ewe type. The specific example under discussion

THE INTERPLAY OF RELEXIFICATION AND LEVELLING

245

further illustrates an early case of levelling where no competing forms are visible in the creole as it is spoken today.7

9.4.

Reflexives

This section shows that the variation between the substratum languages of Haitian with respect to how they encode reflexivity was carried over into the creole through relexification, creating variation in the creole. This variation offers a situation where levelling can take place. In contrast to the case discussed in section 9.3, the variation among competing Haitian forms to encode reflexivity is still visible in the creole, thus making this ongoing process of levelling observable within the Haitian community. The discussion of this case (based on Lefebvre 1998a: 159-183) illustrates the fact that not all cases of levelling take place in the early stage of creole formation. Recall from section 3.3.1.3 that Fongbe has a reflexive anaphor but no BODY-part reflexives. According to Hazoumê (1990: 62), all Gbe languages encode reflexivity by means of a personal pronoun + SELF. In Fongbe, this lexical item is -ɖéè. Kinyalolo (1994) argues that the semantic and distributional properties of this Fongbe lexical item are similar to those of the English -self. For example, in English, pronouns combine with -self (e.g. He washes him self', We wash ourselves; etc.). Similarly, in Fongbe, the strong pronominal forms (but not the syntactic clitics, as is emphasised by both Brousseau 1995a and Kinyalolo 1994) combine with -ɖéè, as is illustrated in (27) in chapter 3, repeated here as (6) for convenience.

(6)

a.

Ni. ná hù ny -ɖéèi. lsg DEF.FUT kill me-SELF 'I will kill myself.'

b. Bàyíi mò é-ɖéèi. FONGBE Bayi see she-SELF 'Bayi saw herself.' (=(45) in Brousseau 1995a)

7 The following question was addressed to me: "[...] why levelling would not have countered the role of relexification in the initial stages of what would become Haitian Creole?" There are two ways in which this question can be interpreted. The first one is whether levelling could take place prior to relexification, thus countering relexification. In this case, levelling would have to operate on the native languages of the early creole population, in this case, the West African languages in presence at the time the creole was formed. Such a possibility was argued against in the discussion of the third scenario. The second interpretation of the above question is whether levelling could start taking place before the relexification of whole lexicons was completed, thus countering the relexification of some lexical entries. In my view, this possibility is a likely one. It is compatible with the fact that, for levelling to take place, the members of a community must communicate in the same language, in this case the incipient creole, consisting of West African lexicons (partially) relexified on the basis of French phonetic matrices. It is also compatible with the first scenario proposed to account for the properties of Haitian yo.

246

ISSUES IN THE STUDY OF PIDGIN AND CREOLE LANGUAGES

Fongbe has no BODY-part reflexives. Recall also from section 3.3.1.3 that both Brousseau (1995a) and Kinyalolo (1994) report that, in Fongbe, the word meaning 'head' does not participate in the expression of reflexivity. Furthermore, they also show that wù, the Fongbe word meaning 'body', does not participate in such constructions either (for an extensive discussion of this point, see Lefebvre 1998 a: 167-171). As is documented in Awoyale (1986), however, BODY-part reflexives are widely used in other Kwa languages. For example, in Yoruba, the word meaning 'body' is ara; this word may appear with personal pronouns, yielding a reflexive interpretation (e.g. ara mi 'body' + 1sg pronoun='myself'). Awoyale (1986) provides similar examples from Igbo, Urhobo, Ebira and Bassa-Nge. Similar data in Efik and Akan are reported by Faltz (1985). Sylvain (1936) discusses a Wolof reflexive formed on bob 'head', and Faltz (1985) mentions a Fula reflexive formed on hooremum (lit.: 'his head') 'himself'. Thus, the substratum languages of Haitian appear to encode reflexivity either by means of a reflexive anaphor or by means of a BODY-part reflexive. This is summarised in (7). (7)

a. b.

reflexive anaphor: BODY-part reflexives:

Gbe languages Yoruba, Igbo, Urhobo, Ebira, Bassa-Nge, Efik, Akan, Wolof, Fula

The hypothesis outlined in section 9.1 predicts that the early Haitian lexicons manifested both a reflexive anaphor and BODY-part reflexives and that this variation was afterwards subjected to levelling. Is there evidence supporting this hypothesis? As has been extensively documented in the literature, reflexivity can be encoded in Haitian in two ways. These are shown in (8). In (8a), the object pronoun is interpreted as a reflexive even though there is no overt reflexive morphology (see Brousseau 1995b; Carden and Stewart 1988; Déchaine and Manfredi 1994; Kinyalolo 1994; Muysken and Smith 1995). In (8b) and (8c), reflexivity is encoded by means of tèt ('head') + pronoun and by means of ko ('body') + pronoun, respectively. The last two forms are also attested in Goodman (1964: 57) and in Sylvain (1936: 63-64). (8)

a. b.

Mweni wè mweni nan glas la. HAITIAN 1sg see me in mirror DEF T saw myself in the mirror.' (=(la) in Brousseau 1995b) Mi ap touye tèt mweni. HAITIAN lsg DEF.FUT kill head me T will kill myself.' (=(2a) in Brousseau 1995b)

THE INTERPLAY OF RELEXIFICATION AND LEVELLING

c.

Lii blese kò 3sg hurt body him 'He hurt himself.'

lii.

247

HAITIAN

(=(6d) in Brousseau 1995b)

Let us assume that all these three possibilities have been produced by relexification. Hence, from this perspective, the phonological representations of the BODY-part reflexives of Haitian would be derived from French tête 'head' and corps 'body', respectively. In French, these two words cannot be assigned a reflexive interpretation. As has been pointed out by Carden and Stewart (1988:11), Goodman (1964:57), Muysken and Smith (1995:276) and Sylvain (1936: 65), among others, there are instances in Old French of the word corps 'body' being used as a reflexive (e.g. Por lor cors deporter 'to amuse themselves', see Muysken and Smith 1995: 276), but there is no evidence showing that this use of the French word corps was available to the creators of Haitian. In fact, grammarians report that the reflexive use of corps had disappeared by the middle of the 16th century (see e.g. Brunot 1905). Therefore, like Goodman (1964: 57), I conclude that it is unlikely that the creators of Haitian were exposed to a variety of French which used BODY-part reflexives. This conclusion challenges those drawn by Chaudenson (1973). As a result of having been created through the process of relexification, these two Haitian lexical entries should have the properties of the corresponding substratum lexical entries. Consequently, in addition to being used to refer to the actual body part or body, they are also used to encode reflexivity, just as in the substratum lexicons that they were relexified from, that is, those listed in (7b).8 Now, what happened in the case of the Gbe speakers who had a reflexive anaphor? In Lefebvre (1998a: 163-167), it is hypothesised that these speakers looked for a form in the superstratum language to relabel the -déè anaphor of their lexicon. It is a well-documented fact in the literature that French does not have a reflexive anaphor like -self in English or like -déè in Fongbe. French expresses reflexivity in other ways (see Lefebvre 1998a: 163-167, and the references therein). Furthermore, the French word même (sometimes unofficially suggested as a potential form for the relexification of the Fongbe reflexive anaphor) did not have the suitable properties, as is extensively argued in Lefebvre (1998a: 164-166). Building on an idea suggested by John Lumsden during a research seminar in 1993, Lefebvre (1998a: 164) proposes that, since the Gbe speakers did not find an appropriate form in the superstratum language to relabel the -déè anaphor, the Haitian lexical entry relexified from -déè was

8

For a different view, see Carden and Stewart (1988: 22) who advocate that BODY-part reflexives might constitute an independent development in creoles.

248

ISSUES IN THE STUDY OF PIDGIN AND CREOLE LANGUAGES

assigned a null form at relabelling, yielding a covert reflexive anaphor in the Haitian lexicon.9 As is argued in Lefebvre (1998a: 166), independent evidence for this claim comes from Berbice Dutch and Gullah whose lexifier languages, Dutch and English, respectively, have a lexical entry meaning '-self'. According to Robertson (1993: 307), Berbice Dutch may express reflexivity by means of a pronoun + SELF. The Berbice Dutch form corresponding to English -self is -selfu, whose phonological form probably derives from Dutch -zelv. Similarly, according to Mufwene (1992: 169), Gullah may express reflexivity by means of a pronoun + SELF. The Gullah form corresponding to English -self is -self, whose phonological form obviously derives from English. Another similar example is provided by data from Saramaccan. The substratum languages of Saramaccan are very much the same as those of Haitian (see Smith 1987). Like Berbice Dutch and Gullah, Saramaccan has an overt -self anaphor: seéi (Veenstra 1996a: 43) in addition to having BODY-part reflexives (Muysken and Smith 1995). Thus, while the creators of Berbice Dutch, Gullah and Saramaccan found appropriate forms in Dutch or English to relabel the lexical entry meaning '-self', the creators of Haitian did not find an appropriate form in French because French does not have such a form. Consequently, they had to assign a phonologically null form to this lexical entry at relabelling. Assuming the representation of a pronoun + SELF in English to be as in (9a), following Chomsky (1981), the representation of a pronoun + SELF in the languages discussed above would be as in (9b).

9

I refer the reader to Lefebvre (1998a: 160-167) for further discussion of this point. For a different view, see Déchaîne and Manfredi (1994).

THE INTERPLAY OF RELEXIFICATION AND LEVELLING

249

In summary, the Gbe speakers in (7a) produced a Haitian lexicon that had a phonologically null reflexive anaphor, accounting for the data in (8a), while the speakers of the other substratum languages in (7b) produced a Haitian lexicon with BODY-part reflexives, accounting for the data in (8b) and (8c). This is summarised in (10). (10) a. b.

reflexive anaphor BODY-part reflexives

GBE (7a) overt -

OTHER KWA (7b) ? 10 +

HAITIAN covert +

The reflexive forms of Haitian constitute a clear example of how relexification proceeds when various substratum lexicons offer different forms to encode a particular notion. The relexification hypothesis thus provides a straightforward account of the fact that we find several reflexive forms in Haitian Creole (and in other creoles as well, see Muysken and Smith 1995). These forms reflect the differences found in the substratum lexicons. This suggests that, in the early creole, there were different Haitian dialects reflecting the differences between the substratum languages. Hence, speakers who had relexified Gbe lexicons would use pronominal forms and a phonologically null anaphor. Speakers who had relexified their lexicons based on other languages such as those mentioned above would use BODY-part reflexives. While communicating with each other, speakers of the early Haitian dialects presumably learned the forms that were originally foreign to them. As a result, it is not rare to find speakers of modern Haitian who use all three reflexive forms. The availability of several forms encoding the same notion constitutes an ideal situation for levelling to take place. Recall from the introduction to this chapter that the process of levelling refers to the reduction of variation between dialects or even idiolects in situations where they are brought together. Hence, it is likely that levelling would operate to reduce the number of forms encoding 10 The literature on reflexives cited earlier only provides examples of BODY-part reflexives. Whether there are Kwa languages other than Gbe that have reflexive anaphors is a question for future research.

250

ISSUES IN THE STUDY OF PIDGIN AND CREOLE LANGUAGES

reflexivity in the various Haitian communities. Evidence that this is the case is provided by Carden and Stewart (1988), who show, on the basis of various types of data (e.g. geographical, historical, etc.), that levelling is now taking place with respect to the ways to encode reflexivity.11 According to them, the beginning of this process goes back to 1790, and it is not yet completed. This shows that, while some cases of levelling are completed in the early stages of the creole (e.g. the case of the plural marker in section 9.3), other cases take much longer to be resolved. This should not come as a surprise if cases of levelling are considered as cases of regular linguistic change which gradually (over generations) take place in communities. 9 . 5 . Demonstrative terms Demonstrative terms constitute yet another case where levelling appears to operate on the output of relexification from various substratum languages. I limit my discussion to the two Haitian deictic lexical entries: sa and sila. Speakers of Haitian appear to divide into two groups with respect to these lexical items. Some have both terms in their lexicon; others only have sa. Most Haitian speakers I worked with on this topic have both terms. The following authors attest the existence of these two forms in the Haitian lexicon that they discuss: d'Ans (1968), DeGraff (1999), Étienne (1974), Faine (1974), Férère (1974), Goodman (1964), Hall (1953), Joseph (1988), Orjala (1970), Peleman (1978), Pompilus (1976), Sylvain (1936), Tinelli (1970), Valdman (1978) and Valdman et al. (1981). The following authors mention speakers who only have the form sa: Valdman (1996) and Vilsaint (1992). I will discuss these two lexicons in turn. 9.5.1 The Haitian lexicon that has both sa and sila Consider the nominal structures of Haitian and Fongbe in (11). The demonstrative terms are glossed as DEM and are translated as 'this/that/these/those' for now. I return below to the patterns of interpretation associated with these terms. (11)

bag

[mwen ø] sa/sila

a

yo

HAITIAN FONGBE

ring me GEN DEM DEF PL 'these/those rings of mine' (in question/that we know of) (=(1) in Lefebvre 1998a: 78) 11 According to these authors, during the levelling period, different geographical areas did not choose the same variant, in such a way that, in this case, levelling appears to also result in dialect differentiation where Port au Prince, the North and the South of Haiti manifest geographical dialectal differences with respect to how reflexivity is now encoded.

THE INTERPLAY OF RELEXIFICATION AND LEVELLING

251

The striking fact about the nominal structures in (11) is that, in both Haitian and Fongbe, a possessor phrase (in the genitive case), a demonstrative term, the definite determiner, and the plural marker may all be present. In both languages, the order of these lexical items is the same. Furthermore, all the "determiners" occur postnominally. This is in direct contrast with French, where the determiners are prenommai and where the three types of determiners are mutually exclusive, as is shown in (12).12

In Lefebvre (1997: 184-193), it is extensively argued that the other syntactic properties of Haitian sa and sila contrast with the properties of the French demonstrative terms (e.g. gender, animacy, syntactic category, distribution, etc.) and that there is no French form that has the same properties as sa and sila in Haitian (a conclusion that challenges that in Chaudenson 1993). 13 Furthermore, in Lefebvre (1997: 193-196), it is extensively argued that the 12 An anonymous reader pointed out the similarity between the position of the possessor phrase in Haitian and the position of the emphatic pronoun in popular French in examples such as la bague à moi là 'the ring of mine there' and raises the question of "[...] whether non standard French structures just had no role to play whatsoever in the development of Haitian Creole". A full discussion of this subject is beyond the scope of this chapter. However, since the question was raised, I will answer it by giving the main points. Although French structures of the type cited above served as a constraint on constituent order in the formation of Haitian Creole nominal structures, as is mentioned in section 9.2 and as is argued in Lumsden (1991)—precisely on the basis of the position of the possessor phrase in Haitian and its contributing languages—, the French possessor phrase and the Haitian one in (11) differ in case. As is extensively argued in Lumsden (1991), whereas the French possessor phrase is marked for objective case à, as in la bague à moi là, the Haitian one in (11) has all the characteristics of a genitive phrase (the genitive case marker is phonologically null in Haitian). According to Lumsden's (1991) analysis, the Haitian possessor phrase in (11) has the same properties as the Fongbe possessor phrase marked for genitive case . The fact that Fongbe marks genitive case rather than objective case is extensively argued in Brousseau and Lumsden (1992). For further discussion of this point, see also Lefebvre (1998a: 101-110). 13 The authors whose work on French deictic terms is cited in Lefebvre (1997) are Brunot (1926), Dees (1971), Grevisse (1975), Guiraud (1966), Lommatzsch (1925), Rosset (1911) and Yvon(1951).

252

ISSUES IN THE STUDY OF PIDGIN AND CREOLE LANGUAGES

details of the syntactic properties of Fongbe and (é)nέ parallel those of Haitian sa and sila, respectively (e.g. gender, animacy, syntactic category, distribution, etc.). On the basis of these facts (and semantic facts discussed below), Lefebvre (1997: 196-198) proposes a relexification account of the two Haitian lexical entries: while the syntactic properties of Haitian sa are derived from Fongbe the phonological representation of Haitian sa is derived from French ça [sa]. Haitian sa and French ça are argued not to have the same syntactic properties. Likewise, while the syntactic properties of Haitian sila are derived from Fongbe (é)nέ, the phonological representation of sila is derived from either one of the following French forms: cela pronounced [sola] or [sla], both yielding sila in Haitian, or celui-là pronounced [sila], also yielding sila in Haitian. This is illustrated in (13).

I now turn to the discussion of the semantics of the lexical items involved. Three semantic patterns have been attested so far for sa and sila. These are given in (14). The first pattern is identified as Gl and it manifests a distal distinction between the two terms. In this grammar, sa is [+ proximate] and sila is [-proximate]. This grammar is attested in the following sources: Goodman

THE INTERPLAY OF RELEXIFICATION AND LEVELLING

253

(1964: 51) 14 and Tinelli (1970: 28) 15 . None of the Haitian speakers I have worked with so far manifest this grammar. The second pattern is identified as G2. In this grammar, while sa is [a proximate], thus a general deictic term, sila is [-proximate]. 16 This pattern is attested in Lefebvre (1997). 17 The third pattern is identified as G3. In this grammar, there is no distal distinction between sa and sila, which can both be used to point at objects that are either close to or far from the speaker. (I will come back to the semantics of this grammar below.) There are speakers with whom I worked who have this grammar. It is also attested in Férère (1974: 103), Joseph (1988: 112), Valdman (1978: 194) and Valdman et al. (1981). (14)

Gl G2 G3

sa [+ proximate] sa [ proximate] sa [a proximate]

sila [-proximate] sila [- proximate] sila [a proximate]

HAITIAN

14 In his comparative grammar of French-based creoles, Goodman (1964: 51) writes: "In Haiti [...], these two meanings ['this' and 'that'] are distinguished formally, sa 'this' and sila 'that'." 15 Tinelli (1970: 194) attests two grammars: "sa is sometimes replaced by sila, with no syntactic change [...], and no clear modification of meaning, except for some speakers who distinguish the remote deictic sila from the proximate sa [...]." 16 This pattern is similar to the first one in the sense that sila is [- proximate]. It differs from the first one in the sense that sa is [a proximate] rather than [+ proximate]. There is independent evidence that sa is a general deictic term for some speakers. Although sa is glossed as 'this' at the beginning of Hall's (1953: 29) book, the 195 cases of this lexical item found in his grammar and Haitian texts are translated either as 'this' or 'these' (57 occurrences) or as 'that' or 'those' (138 occurrences), which shows that sa is a general deictic term in this source. For similar facts, see d'Ans (1968), Peleman (1978) and Sylvain (1936). This argues that the analysis of sa as a general deictic term is common in Haitian. 17 Although no other author has explicitly reported a variety of Haitian where sa is [a proximate] and sila is [-proximate], some Haitian sources provide evidence that suggests such a distinction. For example, in Sylvain (1936), all occurrences of sila are translated by a French deictic term + là: Lõžé sila-yo k'trò bã 'allongez-moi ceux-là qui sont trop loin'/'stretch (give me) those ones which are too far for me' (p. 58); Sila-a sé frè-m 'celui-là est mon frère'/'that one is my brother' (p. 108). In contrast, sa is translated by any of the French deictic terms: Burik-sa-a move 'cet âne est méchant, cet âne-ci est méchant'/'this/that//this donkey is nasty' (p. 56); Sa-a gu 'celui-ci est agréable au goût'/'this one is pleasant to taste' (p. 57); Tut sa-a pa bõ 'celui-là n'est pas bon en entier'/'that one is not completely good' (p. 76). The following example shows that sa may also be used to point at something that is far: M'pito sa k'pi lwe-a 'je préfère celui qui est plus loin'/'I prefer the one that is farther' (p. 149). I take Sylvain's (1936) translations to reflect the pattern I describe: while sila is [-proximate], sa is a general deictic term [a proximate]. Étienne's (1974: 154) remark seems to make the same statement: "Le morphème /sila/ comporte une indication d'éloignement [...]. Il s'oppose à /sa/ [...] comme démonstratif locatif. [...] Les Haïtiens du Nord semblent faire de /sa/ un double emploi: celui de démonstratif général (proche ou lointain) ou de locatif." [The /sila/ morpheme comprises information of distance [...]. It contrasts with /sa/ [...], a locative demonstrative. [...] Haitians of the North seem to have two uses for /sa/: one is a general demonstrative term (close or far), the other one is a locative.]

254

ISSUES IN THE STUDY OF PIDGIN AND CREOLE LANGUAGES

What is the source of these three semantic patterns associated with sa and sila? Are the French forms from which sa and sila are phonologically derived the source of these three patterns? The forms ça, cela and celui-là are all [a proximate] (see Lefebvre 1997: 192-193 for an extensive discussion of this point, based on literature on 17th- century French. See note 14 of this chapter for a list of authors consulted on the semantics of French deictic terms). So the semantics of these demonstrative terms could at best account for Haitian G3, if Haitian sa and sila in G3 were indeed equivalent in meaning to the French forms ça and cela/celui-là, respectively. I doubt it, however, for reasons I come back to below. Also, as is shown in Lefebvre (1997), there are no demonstrative terms in French which lexically encode the value [-proximate]; hence, the [- proximate] feature of sila in Gl and G2 in (14) has no equivalent in French. Furthermore, French lexically encodes the feature [+ proximate] with the form ceci. In this case, the distal distinctions that are lexically encoded in French are as in (15). (15)

ceci 'this' [+ proximate]

ça/cela/celui-là 'this/that' [a proximate]

FRENCH

The interpretive pattern in (15) does not correspond to any of the patterns found in Haitian (see (14)). I thus conclude that the interpretive patterns associated with Haitian sa and sila in (14) are not derived from French. By hypothesis, the semantics of sa and sila come from the substratum languages. (The phonological forms are obviously derived from the superstratum language.) So let us look at what we find in Fongbe. The interpretive patterns that are attested for (é)16 and (é)nέ are shown in (16). The first grammar (Gl) shows a distal contrast between the two terms. This grammar is attested in Anonymous (1983) and Segurola (1963). I also worked with a speaker who has this distinction. In the second grammar (G2), (é)l5 is a general deictic term ([a proximate]) and (é)né is [-proximate] (see Lefebvre 1997, 1999b). In the third grammar (G3), there is no distal distinction between the two terms.18 (16)

18

Gl (é)lo [+proximate] G2 (é)15 [ proximate] G3 (é)lo [ proximate]

(é)né (é)né (é)né

[-proximate] [- proximate] [a proximate]

FONGBE

Lefebvre (1997) is based on fieldwork done prior to 1992 (G2). Gl and G3 were found after this date during fieldwork done with additional speakers.

THE INTERPLAY OF RELEXIFICATION AND LEVELLING

255

The Fongbe interpretive patterns in (16) parallel the Haitian ones in (14). Is this a coincidence? No, it follows directly from the relexification hypothesis which predicts that both the syntactic and semantic properties of the Haitian lexical entries sa and sila should be provided by the corresponding ones in the substratum languages. Although demonstrative terms in Kwa languages have not been given much attention in the literature, the following interpretive patterns are attested for languages other than Fongbe. The Gl pattern of Fongbe in (16) is also attested for the two Ewe corresponding lexical entries (see Rongier 1979: 521, 1995; Westerman 1930). The G2 pattern of Fongbe in (16) is also attested for Ewe in Lefebvre (1997) and Wallace (1995: 6) and acknowledged by Chris Collins (p.c.). The Gl pattern in (16) is also attested in other Kwa languages such as Ega (see Bolé-Richard 1983: 379), Eotile (see Hérault 1983b: 415) and Krobou (see Mensah 1983: 454). So, presumably, and following the methodology adopted above, the interpretive patterns of the substratum lexicons were carried over into the creole in the way schematised in (17).

The first two grammars identified for Haitian in (14) and Fongbe in (16) involve a distal distinction between the two demonstrative terms. The third one, however, does not. For Haitian, DeGraff (1999, and the references cited therein, e.g. Férère 1974; Joseph 1988; Valdman 1978) claims that sa and sila are interchangeable. As DeGraff (1999) puts it, "spontaneous examples abound where sa and sila are used in nominal phrases referring to objects that are both near and equally distant". This remark, and similar ones cited in DeGraff, suggests that sa and sila are synonyms for they apparently show no difference in meaning. Could it be, however, that in G3, another dimension of deixis is involved? We know from the literature that there are various dimensions to deixis: person, time, place, speaker's point of view, etc. (see e.g. Diessel 1999; Fillmore 1975; Grenoble and Riley 1996; Gundel, Hedberg and Zacharski 1993; Levinson 1983; Maclaran 1980; Piwek and Cremers 1996; etc.). Although I did not investigate the matter, I would not be surprised if, in this case, there was another dimension of deixis involved, and this, for two reasons. The first one is that there are West African languages whose deictic terms have been reported to be interpreted along other dimensions of deixis. For example, Jaggar and Buba (1994) have argued that, in Hausa, deictic terms do not always manifest a distal

256

ISSUES IN THE STUDY OF PIDGIN AND CREOLE LANGUAGES

distinction (in much the same way as G3 in (14) and (16)) and that they can be interpreted along non-spatial deixis dimensions (anaphoric and symbolic). The second reason is that, as is discussed in Lefebvre (1998a: 213-217), both Fongbe and Haitian Creole (but not French) have a series of markers that give the point of view of the speaker on the proposition. Could it be that deictic terms are to the nominal structure what these markers are to the clause structure? I submit this as a promising topic for future research. 9.5.2 The Haitian lexicon that has sa but not sila Speakers who have sa but not sila in their lexicon are attested in the literature (e.g. Valdman 1996; Vilsaint 1992). There is one speaker with whom I worked who has only sa. Are there substratum languages of Haitian whose lexicons have only one demonstrative term? Strikingly enough, in the Atlas des langues kwa de Côte d'Ivoire (Hérault (ed.) 1983), several such languages are identified. For example, for Abidji (Tresbarats 1983: 62-63), Adioukrou (Hérault 1983a: 145) and Alladian (Duponchel and Mel 1983:204), only one demonstrative term is reported. Presumably, the speakers of the substratum languages of Haitian who had a lexicon containing only one demonstrative term relexified this term on the basis of the French form ça, yielding sa in Haitian Creole. Consequently, in the early creole, in addition to the lexicon with the two demonstrative terms showing the three interpretive patterns in (14), there was also a lexicon with only one form used as a general deictic term. This is schematically represented in (18). (18)

Haitian lexicon 1 Haitian lexicon 2

sa sa

sila

(three interpretive patterns) (general deictic term)

In this kind of situation, we expect dialect levelling to take place. Is there evidence from the literature that levelling is taking place in this area of the Haitian lexicon? 9.5.3 Levelling Recall from section 9.1 that, in the process of dialect levelling, different communities may arrive at a somewhat different compromise. There is evidence from the literature that different Haitian communities have arrived at different compromises with respect to the hypothesised early Haitian lexicons schematically represented in (18). For example, in his dialect survey of Haitian Creole, Orjala (1970: 117) points out that all dialects have sa and that the northern and central dialects also have sila. This statement is corroborated by Joseph (1988: 112): the form sila "[...] se rencontre surtout dans le nord d'Haïti

THE INTERPLAY OF RELEXIFICATION AND LEVELLING

257

et dans le dialecte de certaines régions de l'Artibonite [...]". [is mostly found in the North of Haiti and in the dialect of some areas of the Artibonite region]. Valdman (1978: 194) also points out that sa and sila constitute dialectal variants. These observations suggest that the two hypothesised early Haitian lexicons in (18) have been subjected to dialect levelling and that different Haitian communities have arrived at different compromises in the process. Another indication that there might be some levelling taking place with respect to the two lexical items sa and sila is frequency of use. Several authors point out that sa occurs more frequently than sila (e.g. Étienne 1974: 154; Férère 1974: 105; Joseph 1988: 112; Pompilus 1976: 38). Furthermore, while Valdman et al (1981) list both sa and sila, Valdman (1996) only lists sa. These observations suggest that the second Haitian lexicon in (18) might be replacing the first one, at least in some regions of Haiti. Only history will tell if regional dialects will stabilise as they are now or if sila will disappear from all the lexicons of Haitian. The data discussed in this section illustrate the fact that the variation between substratum lexicons may be transferred into the early creole by means of the process of relexification. They further show that this variation may be subjected to levelling, which operates on the output of relexification. 9.6. Conclusion In this chapter, I have shown that relexification of various substratum lexicons on the basis of a single superstratum language provides the speakers of the early creole community with a common vocabulary. However, through this process, individuals bring into the incipient creole idiosyncrasies of their native lexicons; thus, the properties of an early Haitian lexical entry may vary slightly among the speakers of the incipient creole. The three sets of data discussed in this paper show that the bulk of the variation found in the incipient creole is attributable to the variation found in the substratum languages rather than that found in the superstratum language. Other similar cases discussed in Lefebvre (1998a) lead to the same conclusion. Furthermore, this approach is congruent with the claim that the creators of Haitian did not have enough exposure to French, at the time the creole was formed, to acquire the details of the properties of French words, as we saw in section 9.2. Relexification of various substratum lexicons also accounts in a principled way for the observation that different substratum languages may contribute to an incipient creole. On the basis of three sets of Haitian data, I have illustrated how levelling operates on the output of the relexified lexicons, reducing the variation. The role of levelling in creole genesis and development accounts for the fact that the properties of some specific lexical entries of the creole may depart from those of the corresponding lexical entries in the individual substratum languages.

258

ISSUES IN THE STUDY OF PIDGIN AND CREOLE LANGUAGES

The data discussed in this chapter illustrate two different types of levelling with respect to the time span required to complete the change. The first case, discussed in section 9.3, (and other similar cases discussed in Lefebvre 1998a) is hypothesised to be an early case of levelling. The two others (discussed in sections 9.4 and 9.5) have not been completed as yet. This situation raises a number of questions for future research. For example, why are some forms levelled out early, while others take more time, if they at all are levelled out. Are there principles underlying levelling that would make the data predictable in this respect? How does early levelling operate? Finally, the data discussed in this chapter are interesting from another point of view. Singler (1988: 29) advocates the view that "when universals, substrate, and lexifier converge with regard to a given phenomenon, such a phenomenon is more likely to enter the creole grammar than when the sources compete". The data discussed in this chapter constitute clear cases of the sources competing in different ways. First, as was extensively documented above, the superstratum and substratum languages do not have much in common in terms of how they lexically encode the various notions discussed in this chapter. Second, the substratum languages are not homogeneous with respect to the lexical items discussed. Nevertheless, the idiosyncratic properties of the substratum languages have made their way into the creole, as was argued above. This shows that creole lexicons and grammars cannot just be a product of universals.

CHAPTER 10

THE EMERGENCE OF PRODUCTIVE MORPHOLOGY IN CREOLE LANGUAGES THE CASE OF HAITIAN CREOLE*

The most intriguing question about creole languages is without doubt that of how they come about. This chapter addresses the problem of the emergence of productive morphology in these languages on the basis of the case study of Haitian Creole (henceforth HC). The aim of this chapter is threefold: 1) propose methodologies that need to be developed in order to study the morphology of creole languages; 2) present a scenario of the emergence of productive morphology in these languages within the framework of what has come to be known as the relexification account of creole genesis (Lefebvre 1998a, and the references therein)1 ; and 3) address the current issues bearing on the study of the derivational morphology of creple languages on the basis of the data discussed in this chapter. This chapter is organised in the following way: Section 10.1 addresses the problem of identifying affixes that are potentionally native (that is productive) to a given creole and proposes tests designed to identify this potentially producti ve morphology. Section 10.2 addresses the issue of the productivity of the affixes identified as potentionally native to a given creole, since some forms that can be identified as potentially native to a creole on the basis of specific criteria may turn out to be unproductive when tokens of occurrence are taken into * The content of this chapter was originally published in the Yearbook of Morphology (2002) (Lefebvre 2003b). Section 3.4, which had to be removed from the original version due to space limitations, has been put back in the present version. The present version also contains notes that also had to be removed from the original version due to space limitations. The content of this chapter was presented at the Siegen International Workshop on the Phonology and Morphology of Creole Languages, August 2001, under the title "On developing methodologies for studying the morphology of creole languages". I would like to thank the participants in the workshop for their most useful comments and questions. I am grateful to Anne-Marie Brousseau, Marijo Denis, Virginie Loranger, Ingo Plag, Albert Valdman and an anonymous reviewer for their comments on an earlier version. Thanks to Marijo Denis, Renée Lambert and Olivier Tardif, for assisting me in the documentation of the data discussed in this chapter, and to Andrée Bélanger for formating the manuscript. The research underlying this chapter is part of the project "Research builds on research" financed by SSHRCC. 1 This account of creole genesis is based on 25 years of funded (mainly by SSHRCC and FCAR) research that I conducted at UQAM and that involved various teams including collea gues and graduate students. For a history of this research, see the Preface to Lefebvre (1998a).

260

ISSUES IN THE STUDY OF PIDGIN AND CREOLE LANGUAGES

account. Section 10.3 is dedicated to developing methodologies for evaluating the productivity of forms that have been proposed to be productive affixes in the literature. The analyses presented in sections 10.2 and 10.3 provide us with a fairly accurate inventory of the productive morphology of HC. Section 10.4 addresses the question of the emergence of this morphology within the relexification account of creole genesis. Section 10.5 concludes the chapter in addressing the issues that are currently at stake regarding the morphology of creole languages on the basis of the data discussed in the first four sections of the chapter. The research underlying this chapter was conducted within the theoretical framework of generative grammar. Accordingly, affixes are assumed to have syntactic features, selectional properties and semantic features. Following di Sciullo and Williams (1987), the morphological head is assumed to determine the syntactic category of a derived word; this head can either be its base or an affix that it contains. Other theoretical assumptions will be introduced in the text whenever necessary. 10.1. Identifying affixes that are potentially native to a creole language The discrepancy in the numbers of productive derivational affixes proposed by different authors for HC reflects the problem posed by the study of the morphology of creole languages. For example, while d'Ans (1968) claims no derivational affixes at all, Hall (1953) presents a list of over 70. Other authors present an inventory of about ten productive affixes (e.g. Filipovich 1987; Tinelli 1970; Valdman 1978; etc.), and still others an inventory of a lot more affixes (e.g. DeGraff 2001). It is a well known fact that creole languages derive the bulk of their vocabulary from their superstratum language. A word having internal structure in this language may, however, not have internal structure in the creole corresponding word. Consequently, an affix that is native to the superstratum language of a given creole may not have a similar status in the creole. Several authors point out the problem of treating the internal structure of Haitian words on par with French structures (e.g. Brousseau, Filipovich and Lefebvre (henceforth BFL) 1989; Filipovich 1987; Tinelli 1970; Valdman 1978; etc.). Furthermore, how do we know whether the following Haitian words have internal structure: wangatè 'magician', chichadd 'stingy person', avidèyman 'visibly', batisman 'boat, ship', bôlèt Totery', poulèt 'pullet', etc.? Given this situation, tests are needed in order to determine whether a creole word has internal structure and what affixes are potentially native to a creole. In the paragraphs that follow, I review the tests that have been proposed by BFL (1989), using HC, to identify affixes that are potentially native to a given creole. First, a form is a potential native affix in HC if it is found affixed to a base that is foreign to French. For example, the verbalising suffix -e is such an affix,

THE EMERGENCE OF PRODUCTIVE MORPHOLOGY...

261

since it attaches to the noun djòb 'job' (borrowed from English) forming the verb djdb-e 'to work'. The attributive affix -è is also an affix of this type since it attaches to the noun wanga 'fetish' (of West African origin) to derive wanga-t-è 'magician' (where -t- is an epenthetic consonant). Second, a form is a potential native affix in HC if it is used with a base that is derived from a French word that does not take the equivalent affix. For example, while the French inversive prefix dé- only attaches to transitive dynamic verbs, HC de- can attach to both transitive and intransitive verbs, whether dynamic or stative. Thus, HC has de-pasyante 'to get impatient' derived from pasyante 'to be patient', whereas French has im-patienter 'to get impatient'. Similarly, the HC nominalising affix -ay attaches to the base kontre 'to oppose' to form the noun kontr-ay 'opposition'. In French, there is no nominal derived from the verb contrer. Furthermore, the suffixes -wa and -yen in HC, which derive nouns referring to one's place of origin, are used on bases where we do not find them in French. For example, while in Haitian we find Soudan-n-wa 'Sudanese' and Ougand-yen 'Ugandan', in French we find Soudan-ais and Ougand-ais, respectively. Similarly, the attributive suffix -è (from French -eur) derives langajè 'chatterbox' from langaj 'language'. There is no French word derived from langage 'language' using -eur. Finally, the adverbial suffix -man (from French -ment) can attach to a base that is not a word in French; for example, it attaches to the adjective alekout ( ale 'action or result of going') and adjectives from verbal bases (e.g. chire 'to tear, rip' > chire 'torn, ripped') is one such example. As will be shown in 4.1, while this

262

ISSUES IN THE STUDY OF PIDGIN AND CREOLE LANGUAGES

process is very productive in HC, it had ceased to be productive in French by the 16th/ 17th centuries in favor of overt affixes (e.g. Haase 1975). These tests help identify affixes that are potentially native to a given creole. They do not provide absolute results, however, for it could be the case that an affix has to be identified as being native to a given creole without any test identifying it. As will be shown in section 10.2.10, this is the case of the Haitian ordinal suffix. The affixes identified as potentially native to a given creole further need to be argued to be productive, for two words that appear to be morphologically related may turn out to be the sole example of its kind. For example, as has been pointed by Crowley (in press: 11), in rural Bislama, the existence of the pair flot 'to float' and flota 'flotation device on fishing net' could suggest that Bislama has a suffix -a phonologically derived from English -er. Crowley remarks that in Bislama "this affix is still only marginal, as any number of logically possible nouns derived on the same pattern are simply not attested." 10.2. Evaluating the productivity of the affixes identified as potentially native to HC Although there is no consensus in the literature as to how to define and measure morphological productivity (e.g. Aronoff 1976, 1980, 1983; Aronoff and Schvaneveldt 1978; Bauer 2001; Corbin 1987; Hay 2001; Plag 1999; etc.), the most important and rather uncontroversial criterion is that, for a process to be called productive, it must be available to speakers to form new words with sufficient frequency in a predictable manner. The frequency problem has received much attention in the recent literature (e.g. Bauer 2001, Hay 2001, Plag 1999, etc.), and a number of methodological tools have been proposed to solve and objectivise it, such as counting neologisms, counting hapaxes in large corpora, etc. At the time the research on HC morphology was conducted (between 1985 and 1993), large corpora were not available.2 So we had to resort to counting, on the basis of data drawn from dictionaries and from elicitation sessions with native speakers, the number of morphological types supplemented by the number of neologisms involving a given form. A few tokens of a given morphological rule were considered not to be enough for it to qualify as productive. Furthermore, since we were interested in the mental lexicon, that is in the speaker's competence in using the lexical entries of his/her lexicon (including affixes) and of combining them to form words, we wanted to make sure that affixes that were identified as productive were in fact available to 2

Note that Jeff Allen is currently creating Haitian corpora that could eventually constitute basic tools for morphological frequency analyses.

THE EMERGENCE OF PRODUCTIVE MORPHOLOGY...

263

speakers to form new words. As for predictability, the following set of criteria were used. First, particular instances of a given form must share meaning and syntactic features. Second, particular instances of a given form must share a coherent set of selectional properties. Third, the output resulting from the concatenation of a given form with a particular base must be predictable. In this section, each affix identified as potentially native to HC in section 10.1 will be discussed in light of these criteria. The discussion builds mainly on work by Brousseau (1994), BFL (1989), Filipovich (1987) and Lefebvre (1998a). The data are basically drawn from these authors' fieldwork, from Valdman et al.'s (1981) HC dictionary, from Védrine's (1992) dictionary of HC verbs, and from Freeman's (1988) inverse dictionary of HC.3 Due to space limitation, only a few examples of words derived by means of the affixes discussed in this section will be provided. Additional data are easily accessible in the references cited for each affix. 10.2.1 The agentive suffix -è The suffix -è derives agentive nouns from dynamic verbs. The output of this concatenation is a noun with the meaning 'one who VERBs'. (1)

Agentive suffix -è: [V-è]N dechouk-è 'insurgent' konsey-è 'counsellor' rans-è 'joker' vant-è 'braggart'

dechouke konseye ranse vante

'to uproot' 'to counsel' 'to joke' 'tobrag'

(=(2) in BFL 1989)

A sample of Haitian words derived by means of the agentive suffix -è, for which the corresponding French word has a different meaning from the Haitian, or for which there is no corresponding French word, is provided in Lefebvre (1998a: 403). Words derived by means of this affix are numerous (e.g. Freeman 1988; Valdman et al 1981). 10.2.2 The attributive suffix -è There is another suffix -è which is attributive. As is pointed out in BFL (1989), this affix is distinct from the preceding one not only in meaning but also because it attaches to a nominal rather than a verbal base. The derived meaning is 'one who possesses or uses NOUN'. 3

More recent dictionaries such as Freeman and Laguerre (1996) were not available at the time the morphological data were gathered. Although no dictionary will ever be absolutely complete, Valdman et al. (1981) is an extremely reliable source of data. Using this source as our main source has no impact on the global results of this research. For the sake of completeness, however, more dictionaries could be consulted. I leave this task for future research.

264

(2)

ISSUES IN THE STUDY OF PIDGIN AND CREOLE LANGUAGES

Attributive odyans-è langaj-è tafya-t-è wanga-t-è

suffix -è: [N-è]N 'joker' 'chatterbox' 'alcoholic' 'magician'

odyans langaj tafya wanga

'joke' 'language' 'alcoholic beverage' 'charm' (=(3) in BFL 1989)

A sample of Haitian words derived by means of the attributive suffix -è for which there is no French counterpart is available in Lefebvre (1998a: 403-404) and in Valdman (1978: 140). There are numerous Haitian words that are derived by means of this affix (e.g. Freeman 1988; Valdman et al. 1981). 10.2.3 The verbalising suffix -e The suffix -e derives verbs from nominal bases, as shown in (3). (3)

Verbalising suffix -e: [N-e]v betiz-e 'to talk nonsense' kle-t-e 'to lock up' madison-n-e 'to cast a spell' makak-e 'to hit with a stick'

betiz 'nonsense' kle 'key' madison 'spell' makak 'stick' (from Brousseau 1994; BFL 1989)

A large sample of Haitian verbs derived by means of the verbalising suffix -e for which there is no French counterpart or for which the French counterpart has a different meaning may be found in Lefebvre (1998a: 404) and in Valdman (1978: 132). Again, several words derived by means of this affix are listed in Freeman (1988) and in Valdman et al (1981). 10.2.4 Inversive and privative deThe prefix de- is very productive in HC. It may derive new verbs that have a meaning opposite to that of the base form. A few examples are provided in (4). (4)

The prefix de-: [de-V] : inversive meaning deboutonnen 'to unbutton' boutonnen 'to button' dechaje 'to unload, to ejaculate' chaje 'to load' defèt 'to undo, to disconcert, fè 'to do, to make, to get rid of' to produce' defòme 'to deform' fòme 'to form' degonfle 'to deflate, to scatter, gonfle 'to blow up' to make room for' derespekte 'to be disrespectful' respekte 'to respect' (from Valdman et al 1981)

THE EMERGENCE OF PRODUCTIVE MORPHOLOGY...

265

Note that, as is pointed out by Damoiseau (1991: 33), affixation of de- to a verbal base may modify the aspectual properties of the verbal base. For example, while respekte 'to respect' is stative, de-respekte is dynamic. The prefix de- may also derive new verbs that are semantically ambiguous between an inversive and a privative interpretation. For example, debare in (5) can be interpreted either as having an inversive 'to unblock' or a privative 'to remove the blocking material' interpretation. Likewise, debouche 'to open' can be assigned an inversive interpretation with respect to bouche 'to stop up', or it can be assigned a privative interpretation: 'to uncork' versus bouche 'to cork'. In a similar way, and as is discussed in Chaudenson (1996b: 27), the verb degrese is ambiguous between an inversive reading 'to lose weight', and a privative one 'to remove fat' ; and so on and so forth. (5) The prefix de-: [de-V]v: inversive/privative meaning debare 'to clean (remove obstacles)' bare 'to block off, to fence in' debouche 'to uncork, to open (bottle)' bouche 'to stop up, to cork' degrese 'to lose weight, to remove fat' grese 'to grease, to oil' dekloure 'to remove nails' kloure 'to nail' (from Valdman et al 1981) The prefix de- may also participate in the derivation of new verbs that have a privative meaning. These cases involve a nominal base and the verbalising suffix -e, discussed above. The examples in (6) show the verb containing de-, the corresponding nominal base; they also show that the denominal verb (without de-) formed by affixation of -e is unattested. (6) The prefix de-: privative meaning debagaje 'to move bagay (change residence)' dechouke 'to uproot, to get chouk someone fired, to remove from office, to plow' degagannen 'to cut the throat of, gagann to overcharge' degouse 'to shell, to husk' gous dekreta 'to remove the crest' krèt dekwennen 'to skin' kwenn demachwele 'to smash one's machwè face in' demwèle 'to remove the marrow"mwèl 4

*kreta is also discussed in DeGraff (2001: 80).

'thing'

*bagaje

'stump'

*chouke

'throat'

*gaganne

'pod, clove' 'cockcomb' 'rind' 'jaw'

*gouse *kreta4 *kwennen *machwèle

'marrow'

*mwèle

266

ISSUES IN THE STUDY OF PIDGIN AND CREOLE LANGUAGES

depaye

'to take apart a woven straw object' deplancheye 'to remove the floor boards' devantre 'to disembowel' dezose 'to debone'

pay

'straw, chaff'

*paye

planche 'wooden floor' *plancheye vont zo

'belly, stomach' *vantre 'bone' *zose

Derived words of the type of those in (6) are known as parasynthetics.5 With Brousseau (1994), I assume that the inversive or privative interpretation of words containing the prefix de- follows from the semantics of the verbs it attaches to. 10.2.5 The diminutive prefix tiThe Haitian noun pitit 'child' (Valdman et al 1981) has a reduced form ti 'kid' (Valdman 1996). The adjective piti 'small' has a reduced form ti 'small, little' (Valdman et al. 1981). The short form of both words serves as a prefix that attaches to nominal bases. Its meaning with animate bases is 'a young NOUN', as is illustrated in (7) and (8). (7)

ti-chat ti-chen ti-towo

(8)

ti-nèg ti-moun ti-bèt ti-gason ti-medam

'kitten' 'puppy' 'calf

chat chen towo

'young man' nèg 'child' moun 'insect, bug' bet 'boy' gason 'young ladies, girls' medam (from Valdman

'cat' 'dog' 'bull' (from Valdman et al 1981) 'man' 'person' 'animal' 'male (man, boy), waiter' 'women, ladies' 1996; Valdman et al 1981)

With inanimate nominal bases, the prefix ti- has the meaning 'a small NOUN' (e.g. Filipovich 1987: 34-5; Sylvain 1936: 32; Valdman et al 1981). This is illustrated in (9). (9)

5

a.

ti-flè 'small flower' ti-chante 'small song'

b.

ti-wdch 'pebble' ti-pyebwa 'bush'

fiè chante

'flower' 'song' (from Sylvain 1936: 32) wòch 'stone' pyebwa 'tree' (from Filipovich 1987: 34-5)

Various analyses of parasynthetics have been proposed in the literature. Due to space limitations, the possible derivations are not discussed here.

THE EMERGENCE OF PRODUCTIVE MORPHOLOGY...

c.

ti-kiyè ti-van

'teaspoon' 'breeze'

kiyè van

267

'spoon' 'wind' (from Valdman et al 1981)

Many Haitian words that are derived by means of the prefix ti- have no corresponding French words. A sample of these can be found in Lefebvre (1998a: 403-4). Since the prefix ti- is homophonous with the noun ti and the reduced form of the adjective ti, are there arguments in favor of the affixal status of ti- in the above examples? Filipovich (1987: 33-4) argues against an Adj + N analysis of the forms in (9) on the following grounds. First, the position of ti- contrasts with that of piti. While ti- always precedes the noun (a fact also mentioned in Valdman et al. 1981), piti never occurs in this position, as is shown in (10a). While piti always occurs as a stative predicate, ti never occurs with this function, as is shown in (10b). (10)

a.

*piti machin yo small car PL [Lit.: 'the small cars']

b.

Machin yo pitil*ti car PL be.small 'The cars are small.' (from Filipovich 1987: 33-4)

The complementary distribution of ti and piti argues against the adjectival status of ti in the examples in (9). Second, the nouns in (9) may be modified by the adjective gran 'big', as in se youn gran ti-wòch 'It is a big pebble.' If ti were an adjective in the examples in (9), we would not expect the modifier gran to occur with it. If ti is an affix, however, the word that contains it can be modified by an adjective meaning 'big' yielding the interpretation 'for a small x, it is big', without there being any semantic contradiction.6 This argues in favour of the affixal status of ti- in (9). As for the nouns in (7), they can also be modified by an adjective meaning 'big', as in yon gran ti-moun 'a big child', thus arguing against an adjectival analysis of ti- in these cases as well. Furthermore, on the basis of tests that distinguish between derivation and compounding, ti- can be argued to be an affix rather than part of a compound. First, in contrast to compound words, affixed words show more productivity and more semantic regularity (e.g. Allen 1978; Aronoff 1976; di Sciullo and Williams 1987). Whereas the meaning of a compound cannot always be derived from its compositional meaning, the meaning of a productively derived word is predictable from the meaning of its affix. As is shown in (7) and (8), the 6 [Marijo Denis (p.c.) remarks that the same facts can be observed in French. While it is possible to say 'une grande maisonnette' it is not possible to say une grande petite maison.]

268

ISSUES IN THE STUDY OF PIDGIN AND CREOLE LANGUAGES

meaning of the words containing ti- is predictable from the meaning of the affix. Although there are words containing ti- that manifest idiosyncrasies, as will be seen in section 10.5.4, the properties of ti- in (7) and (8), are those of a productive affix. Second, while it is not always possible to predict the actual combinations of words into compounds, an affix systematically attaches to all words that meet its subcategorisation requirements. As we saw above, tiattaches in a regular way to nominal bases. I thus conclude that words containing ti- result from morphological derivation rather than from compounding.7 10.2.6 The nominalising suffix -ay The suffix -ay derives nouns from two-place verbal predicates. The nouns so derived refer either to the action or to the result of the action denoted by the verbs. (11)

Nominalising suffix -ay: [V-ay]N kontr-ay 'opposition' bwot-ay 'moving' dechouk-ay 'overthrow' kapon-ay 'intimidation'

kontre bwòte dechouke kapone

'to oppose' 'to move' 'to uproot' 'to intimidate' (=(6) in BFL 1989)

Although there is variation between speakers as to the properties of -ay (see Lefebvre 1998a: 306-307), none of the speakers consulted accept a nominal structure where both arguments are realised in the syntax. Hence *dechoukay presidan pèp ayisyen an with the reading 'the overthrow of the president by the Haitian population' is ruled out by both groups of speakers. The following two structures are found: dechouk-ay presidan an 'the overthrow of the president' or dechouk-ay pèp ayisyen an 'the overthrow by the Haitian population'. A sample of Haitian words derived by means of the suffix -ay for which there are no French corresponding words is provided in Lefebvre (1998a: 406) and in Valdman (1978: 136). Numerous Haitian words containing the nominalising suffix -ay are found in the inverse dictionary (Freeman 1988), and in Valdman et al (1981). 10.2.7 Morphological conversion Morphological conversion, a process that has been extensively discussed in the literature on Haitian (see Fauchois 1983; Filipovich 1987; Hall 1966; Sylvain 1936; Tinelli 1970; etc.), derives nouns and adjectives/participles from verbs. Nouns that are derived from verbs by conversion refer either to the action or to 7

For a compounding analysis of the data in (7) and (8), see Valdman (1978: 155).

THE EMERGENCE OF PRODUCTIVE MORPHOLOGY...

269

the result of the action denoted by the verbal base. Some examples are provided in (12). (12)

Nominalising conversion: [V]N ale 'action or result of going' sòti 'action or result of going out' vini 'action or result of coming' rive 'action or result of arriving' tonbe 'action or result of falling' desann 'action or result of descending'

ale söti vini rive tonbe desann

'to go' 'to exit' 'to come' 'to arrive' 'to fall' 'to descend' (from BFL 1989)

As is shown in BFL (1989), this process applies only to verbal bases having only one argument: unaccusative verbs (e.g. ale 'to go'), unergative verbs (e.g. krache 'to spit') and the intransitive versions of verbs which show a transitive/ intransitive alternation (e.g. bwdte 'to move'). The derived noun has the predicate argument structure of the verbal base. In all cases, the single argument of the verb may be expressed in the syntax, as is illustrated in (13). (13)

a.

ale Jan departure John 'John's leaving'

an DEF

b.

bwdte Mari a moving Mary DEF 'Mary's moving' (=(8a) in Lefebvre 1998a: 308)

Nominal conversion is very productive in Haitian. Lefebvre (1998a: 406) provides a sample of nouns derived by conversion and shows that only a few of them have a corresponding French noun derived by this process, even when 17th/18th century French data are considered. Védrine (1992) lists no fewer than sixty verbs that may be nominalised by means of this process. This process is in complementary distribution with the derivational process involving the suffix -ay (see (11)). (For a discussion of the pertinent facts, see BFL 1989: 15).8 Morphological conversion also derives adjectival and participial forms from a verbal base.9 Védrine (1992: 167-8) lists 254 verbs that can be used as attributives. A few examples are reproduced in (14). 8 Conversion is a process that manifests the properties of affixes. It is different from lexical entries created by relisting where each relisted entry needs to be specified for semantic and morphosyntactic properties because these are not always predictable (for discussion see Lieber 1992). In HC there are a few nouns that are created by relisting, antre (
270

(14)

ISSUES IN THE STUDY OF PIDGIN AND CREOLE LANGUAGES

Adjectival/participial conversion: [[V]]A/PA abandonnen 'abandoned' abandonnen chire 'torn/ripped' chire boure 'stuffed' boure kwit 'cooked' kwit

'to abandon' 'to tear/rip' 'to stuff' 'to cook'

These derived forms are found in verbal and adjectival passive clauses, as illustrated in (15a), or in noun phrases, as in (15b). (15)

a.

Vyann nan kwit. meat DEF cooked 'The meat is cooked.' or 'The meat has been cooked.'

b.

vyann kwit la meat cooked DEF 'the cooked meat' (=(9) in Brousseau 1994)

This derivational process applies to all verbs involving a change of state (e.g. (14)). It does not apply to verbs involving a change of location such as ale 'go', pouse 'push', tire 'pull', nor to verbs of movement such as tonbe 'fall', sote 'jump', naje 'swim'. I will assume without further discussion that both cases of morphological conversion can be represented in a unified way as [[V]]N/A. 10.2.8 The adverbial suffix -man The adverbial suffix -man selects adjectival bases as is examplified in (16). (16)

-man: [A-man]Adv 'not as well, rather badly' mal 'badly' 'equally, the same' menm 'same' fini 'completed' 'completely, quite' (from Freeman and Laguerre 1996) b. angran-man 'haughtily, imperiously' angran 'arrogant, haughty' mal-man 'not so well, rather badly' mal 'badly' menm-man 'equally, the same' menm 'same' (from Valdman et al 1981) banda-man 'elegantly' banda 'elegant' fin-man 'completely' fin 'complete' (from Valdman 1978: 142) c. avidèy-man 'visibly' avidèy 'visible' alekout-man 'attentively' alekout 'attentive' bosal-man 'wildly' bosal 'wild' alèz-man 'comfortably' alèz 'comfortable' bòzò-man 'elegantly' bòzò 'elegant' (A.-M. Brousseau's field notes) Adverbial suffix a. mal-man menm-man fini-man

THE EMERGENCE OF PRODUCTIVE MORPHOLOGY...

271

There are no French adverbs corresponding to the Haitian examples in (16).10 10.2.9 The place of origin/residence suffixes -wa and -yen The suffixes -wa and -yen derive nouns from nouns. The input noun is the name of a place and the output noun refers to a person who comes from that place, as in (17). (17)

a. Place of origin/residence suffix -wa: [N-wa]N Kap-wa 'from the Cape' Kap 'Cape' Petyonvil-wa 'from Pétionville' Petyonvil 'Pétionville' Senmak-wa 'from Saint-Marc' Senmak 'Saint-Marc' Soudan-n-wa 'from Sudan' Soudan 'Sudan' Tchad-wa 'from Chad' Tchad 'Chad' (A.-M. Brousseau's field notes) b. Place of origin/residence suffix -yen: [N-yen]N Jakmèl-yen 'from Jacmel' Jakmèl 'Jacmel' Tibè-yen 'from Tibet' Tibè 'Tibet' Gan-yen 'from Ghana' Gana 'Ghana' Ougand-yen 'from Uganda' Ouganda 'Uganda' Pakistan-n-yen 'from Pakistan' Pakistan 'Pakistan' Ka-yen 'from Les Cailles' Okay 'Les Cailles' (A.-M. Brousseau's field notes)

The principles governing the choice between the two affixes -wa and -yen appear to be as obscure as those governing the choice between the French affixes which perform the same function. The question of how stable these derivations are across Haitian speakers is a question for future research. 10.2.10 The ordinal suffix -yèm This affix -yèm derives ordinal numbers from cardinal ones. (18)

santyèm katriyèm twazyèm

'hundredth' 'fourth' 'third'

(from Valdman et al 1981)

Although the morphophonemics of the Haitian examples in (18) is parallel to that of the corresponding French forms (/santyèm/, /katriyèm/, /trazyèm/), Brousseau (p.c.) remarks that it is unlikely that the creators of Haitian learned the ordinal 10

Orjala (1970: 155) also notes what he refers to as 'hypercorrected' forms, where the adverbial suffix -man is attached to bases that are already adverbial: (i) toujou-man (
272

ISSUES IN THE STUDY OF PIDGIN AND CREOLE LANGUAGES

numbers of French one by one. Brousseau thus proposes that, by virtue of its productivity, -yèm should be analysed as part of the inventory of Haitian affixes, even though it has not been identified as potentially native to Haitian on the basis of the criteria discussed in section 10.1. An argument supporting her claim is provided by data that she collected, which bear on the morphophonemics of the formation of cardinal numbers out of ordinal ones for some Haitian speakers. Brousseau's informants have the forms in (19), where the morphophonemics of the Haitian derived forms differs from that of the French corresponding ones. Compare the forms in (19) with those in (18). (19)

Ordinal suffix san-yèm mil-yèm kat-yèm twa-yèm

-yèm: [cardinal number -yèm] 'hundredth' san 'hundred' 'thousandth' mil 'thousand' 'fourth' kat 'four' 'third' twa 'three' (A.-M. Brousseau's field notes)

10.2.11 Summary All the forms discussed in this section have the properties of productive affixes. Particular instances of each form share meaning, syntactic features and selectional restrictions. The outputs resulting from the morphological rules are predictable. The forms discussed in this section all occur with some frequency. Finally, a careful examination of Valdman et al.'s (1981) dictionary provides quite a large number of Haitian derived words which do not correspond to French derived ones, as well as Haitian derived words which have a different meaning than their French counterparts. On all these criteria, the affixes identified in section 10.1 as potentially native to HC turn out to indeed be genuine productive affixes in this language. The morphological forms discussed in this section exhaust the list of derivational affixes identified as native to, and productive in, HC within the context of the UQAM HC projects. The inventory of the productive affixes of Haitian, based on the analyses presented in this section, comprises eleven affixes. These are listed in Table 10.1.

THE EMERGENCE OF PRODUCTIVE MORPHOLOGY...

273

Table 10.1. The inventory of HC productive affixes -è Agentive suffix base V output N Attributive suffix -è base N output N Verbalising suffix -e base N output V Inversive/privative prefix debase V output V Diminutive prefix tibase N output N Nominalising suffix -ay base V output N Conversion base V output N/A -man Adverbial suffix base A output Adv Place of origin/residence suffixes -wal-yen base N output N -yèm Ordinal suffix base Q output A

(adapted from Lefebvre 1998a: 312) 1 0 . 3 . Evaluating other proposals in the literature DeGraff (2001) claims that there are at least eight additional derivational affixes that are productive in HC. He identifies the prefix en-, and the suffixes -ab, -add, -èt, -man, -syon, -es and -te. In this section, I discuss in detail the properties and the distribution of three of four of these forms: en-, -èt, -man and -add. [The other forms are discussed in Denis (in preparation).] It is shown that these forms do not present the characteristics of productive affixes. Therefore, the words that these forms are part of are best analysed as being listed in the mental lexicon rather than as being productively derived. 10.3.1 The hypothesised inversive prefix enOn the basis of the existence of the word enkoutab 'foolhardy' that he relates to koute 'to listen', DeGraff (2001: 59) concludes that en- is a productive prefix in HC. The properties of en- are not specified. Furthermore, according to the principles of morphological analysis, en-koutab cannot be claimed to be derived directly from koute. It should be derived from the unattested form *koutab, itself derived from koute. So, the sole example presented by DeGraff to support his proposal that en- is a productive affix in Haitian is not well chosen.

274

ISSUES IN THE STUDY OF PIDGIN AND CREOLE LANGUAGES

In order to test the productivity of this hypothesised affix, a corpus of words containing en- was constituted. In a first step, all the words beginning with en- in Valdman et al. (1981) were considered. From this preliminary list, words in which en- is clearly not separable from the rest, such as endistri 'industry' or endyen 'indian', were removed. This left us with the 20 lexical items that are listed in (20), (21) and (22). Possible bases from which the words beginning in en- could be derived from are provided in the right column. Unless otherwise specified, the data in (20)-(22) are from Valdman et al. (1981). Other authors were consulted when the possible base word was not listed in Valdman et al. Potential bases that are not listed anywhere in available sources are identified by an asterisk. Lexical items cited from Freeman (1988) in (20) appear without glosses because this dictionary does not provide translations. Out of the 20 pairs of en- derivatives and their bases, 13 correspond to pairs that are also found in French (e.g. H. depandanslendepandans; Fr. dépendancelindépendance). These pairs, listed in (20), do not argue for or against the productivity of the form en- in HC. (20) endepandans 'independence' endesan 'immodest, lewd' endijesyon 'indigestion' enjis enjistis enkapab enkonsyan enkredil enkwayan enposib enpridan enpridans enpasyans

depandans (Freeman 1988) desan 'decent' dijesyon (Faine 1974) 'digestion' 'unfair' jis 'fair' 'unjust action' jistis 'justice' 'incompetent' kapab 'to be able, can' 'unprincipled' konsyan (Freeman 1988) 'incredulous (sceptical)' kredil (Faine 1974) 'credulous' 'unbeliever' kwayan 'believer' 'impossible' posib 'possible' 'imprudent' pridan 'prudent' 'imprudence' pridans 'prudence' 'impatience' pasyans 'patience'

In three cases the semantic relationship between the hypothesised derived word and its possible base is not straightforward, as is shown in (21). (21)

endispoze

'to faint, dispoze 'disposed to make nauseous' (inclined)' endispozisyon 'fainting spell, nausea' dispozisyon 'disposition (temperament)' envizib 'celestial beings vizib 'visible' (voodoo), invisible' (from Valdman et al. 1981)

THE EMERGENCE OF PRODUCTIVE MORPHOLOGY...

275

As we saw earlier, productive affixes derive words whose semantics is predictable from the meaning of its parts. This is not the case of the words in (21). In the last four cases, there is no base onto which the hypothesised affix en- can be shown to attach to. These are listed in (22). (22)

enkoutab

'stubborn, headstrong' (attributive) *koutab 'stubbornness' (noun) enkwayab 'unbelievable' *kwayab ensipotab 'unbearable' * sipotab ensiyifyan 'stupid, foolish' *siyifyan (from Valdman et al. 1981)

Moreover, note that enkoutab in (22) is not semantically related to koute 'to listen' since, as an attributive, Haitian enkoutab means 'stubborn, headstrong', and as a nominal, 'stubbornness'. This contrasts with its French countepart inécoutable which means 'impossible to listen to'. It thus appears that en- is not a productive affix of HC. 10.3.2 The hypothesised suffix -ét On the basis of the pair of words bòl/bòlèt, DeGraff (2001: 59-61) concludes that there is a derivational process in HC that involves a suffix -èt. No other examples are provided. The selectional properties of this alleged suffix are not specified, nor are its semantic properties. Let us look at the Haitian words ending in -èt on the basis of a corpus constituted in the following way. As a first step, all the words ending in -èt were pulled out of Freeman's (1988) inverse Haitian dictionary. Due to the fact that the inverse dictionary does not provide translations, translations were sought in Valdman et al.'s dictionary. The words that are not in Valdman et al.'s dictionary were eliminated. Hypothesised derived words for which no possible bases could be found in Valdman et al.'s (1981) were removed from the remaining list. The verbal idiosyncrasy chatouyèt 'to tickle', and the nominal sirèt 'hard candy' both exemplify this situation. The remaining list of lexical items ending in -èt was then divided in terms of the syntactic features of the base: nominal and verbal. I begin the discussion with words ending in -èt that have a potential nominal base. Consider the nouns in (23). There is no semantic relationship between the potential bases and the presumed derived words. This indicates that the form -èt in the words in (23) does not have a meaning of its own. (23)

bichèt klochèt kouchèt kwochèt

'winnowing tray' 'kind of flower' 'diaper' 'hook', 'tripping'

bich kloch kouch kwoche

'chunk' 'bell' 'layer' ' coachman '/koche 'scratch' (from Valdman et al. 1981)

276

ISSUES IN THE STUDY OF PIDGIN AND CREOLE LANGUAGES

Since -èt does not appear to have a meaning of its own in the words in (23), it cannot be claimed to be a productive affix on the basis of these words. In (24), there is some semantic relationship between the base and the hypothesised derived word. This relationship is, however, not straightforward nor regular. (24) tètbòchèt 'thin short hair' bwòs [hair cut in form of a brush] choukèt 'short section of tree chouk (used as stool), short person' holet 'lottery' bòl

poulèt kamyonèt chemizet boulèt salopèt

'pullet' 'pick-up truck, van, station wagon (used for public transportation' 'undershirt, t-shirt' 'ball' (meatball, fishball, cannonball) 'slovenly woman'

'brush' 'stump'

poul kamyon

'lottery number formed by two identical digits (DeGraff,2001:60 Peleman 1986: 45) 'chicken' 'truck'

chemiz boul

'shirt' 'ball (sphere, toy)'

salòp

'immoral person, coward, slovenly' gôjèt 'swallow (sip)' gòj 'throat' rakèt 'type of cactus' rak 'scrub' kaskèt 'cap' kas 'hardhat, helmet' pantalèt '(women's) underpants' pantalon 'trousers' fiyèt (lalo) 'folk character fi 'woman' (witch/powerful armed Duvalier partisan (female)) malèt 'suitcase' mal 'trunck' langèt 'clitoris' lang 'tongue' grandèt 'adolescent' grand 'old' (from Valdman et al 1981) Recall from the introduction to section 10.2 that the output resulting from the concatenation of a given form with a particular base must be predictable. This is because productive affixes have a specific meaning that manifests itself in most instances where they occur. So, the question arises as to whether a meaning can be identified for -èt in the words listed in (24). On the basis of a pair like poul/poulèt, one can hypothesise that -èt would have a 'diminutive' interpretation. However, tèt bòchèt is not a small bwôs, choukèt is not a small chouk, and bòlèt is not a small bòl. Even poulet is not just a small or baby poul; rather it is a 'FEMALE baby poul'. Likewise, kamyonèt means more than 'a small kamyon'. The hypothesised derived words chemizèt and boulèt also have more meaning

THE EMERGENCE OF PRODUCTIVE MORPHOLOGY...

277

than the combination of their base and -èt would predict. The word salopèt refers to a woman that has the property of the hypothesised base. A gôjèt is not a small gòj, nor a type of goj. The remaining examples in (24) all present some idiosyncrasy. On the basis of this discussion, I conclude that the form -èt in the words in (24) is not a diminutive suffix. It does not seem possible to associate it to any other meaning that would predict the meaning of the words it is part of. I thus conclude that the form -èt occurring in (24) does not qualify as a productive affix. The examples in (25) consist of pairs where the words ending with -èt and those without -èt are synonymous. (25)

gòdèt kòlèt

'mug, cup' 'collar'

gode 'mug, cup' kòl(e) 'collar' (from Valdman et al 1981)

Thus, again, the form -èt occurring in (25) does not qualify as a productive affix. The words in (26) present cases where the hypothesised morphological derivation is either morphologically or phonologically not straighforward. For example, in the first example, affixation of -èt to kòk yields the unattested word *kôk-èt. Likewise, affixation of -èt to kokoye yields the unattested word *kokoyèt. Instead, we find kokonèt, possibly derived from koko 'genitals (female)/cocoon' (see Valdman et al. 1981) and -èt, (and the epenthetic consonant -n-). If this derivation were the correct one, the semantics of koko 'female genitals, cocoon' and that of kokonèt 'coconut cookie' would be related. From their translation, these two words do not appear to be semantically related. As for the second example, affixation of -èt predicts the unattested form *kasèt. The observed form is kaskèt [kaskt]. If this word has simply been incorporated into HC directly from French casquette (pronounced [kaskt]), the observed Haitian phonological form follows in a straightforward way. Finally, the last word, pantalèt is idiosyncratic for, if it were derived from pantalon, it would have the unattested form *pantalonèt, rather than the attested one pantalèt. (26)

kokonèt

'coconut cookie'

kaskèt pantalèt

'cap' '(women's) underpants'

kôk

'coconut' short form of kokoye 'coconut' kas 'hardhat, helmet' pantalon 'trousers' (from Valdman et al 1981)

The form -èt occurring in (26) thus does not qualify as a productive affix. Finally, chofrèt in (27a) is not derived from the unattested base *chofr- plus -èt. Rather, it consists of a combination of cho 'hot' +frèt 'cold', yielding chofrèt

278

ISSUES IN THE STUDY OF PIDGIN AND CREOLE LANGUAGES

'chill'. Likewise, bwatèt 'skull' in (27b) is a frozen compound comprised of bwat 'box' and tèt 'head' (A.-M. Brousseau p.c.). (27)

a. b.

chofrèt bwatèt

'chill' 'skull'

*chofr-èt *bwat-èt

(from Valdman et al 1981)

It thus appears that the form -èt occurring with nominal bases does not have the properties of a productive affix in HC. I now turn to the discussion of words ending in -èt for which the hypothesised base is verbal. These are presented in (28) and (29). In (28), there is no semantic relationship between the available verbal base and the hypothesised derived noun. (28)

k(w)ochèt

1. 'hook' 2. 'tripping'

kòche

1.'to butcher' 2. 'to scratch' 3. 'to cut oneself' 4. 'to beat (in a game)' (from Valdman et al 1981)

This piece of data does not argue in favor of the morphological productivity of the form -èt. In (29), there is a semantic relationship between the verbal base and the corresponding noun ending in -èt. In all cases, the noun containing -èt is a cognate object of the verbal base. (29)

digèt devinèt sousèt pikèt souflèt

'stick/goad' 'riddle' 'sucking, pacifier' 'peg, stake' 'whistle'

dige devine souse pike soufle

'to goad' 'to guess' 'to suck' 'to sting, to prick, to stab' 'to whistle' (from Valdman et al 1981)

Except for the pair digèt/dige, the pairs of words in (29) all correspond to pairs that are also found in French. This state of affairs raises the question of whether the nouns ending in -èt in (29) are productively derived by means of -èt or whether they have been incorporated as such into the Haitian lexicon. If the Haitian words in (29) do have internal structure, it should be possible to find other words so formed that meet at least one of the criteria discussed in section 1. Having gone through all the Haitian words ending in -èt in Freeman's (1988) inverse dictionary, I did not find any such pairs, except for the pair digètldige. Finally, for -èt to be analysable as a productive affix of Haitian, it should be possible to identify the type of verbs that are selected by this affix. Since the nouns in (29) are all cognate objects, a possible class of verbs to look for would

THE EMERGENCE OF PRODUCTIVE MORPHOLOGY...

279

be cognate object verbs. However, verbs that typically take a cognate object in related languages are intransitive in HC. These are listed in (30). (30)

tete 'to suckle' pise 'to pee' sifle touse 'to cough'

kwè 'to believe' 'to whistle' genyen 'to win'

(from Lefebvre 1998a: 281)

It thus appears that the form -èt occurring in (29) does not have the properties of a productive affix. On the basis of the above discussion, I conclude that it has not yet been proven that -èt is a productive affix of Haitian. 10.3.3 The hypothesised nominalising suffix -man On the basis of the pair of words koze 'to talk' and kozman 'talk, gossip', DeGraff (2001: 60) concludes that there is a productive nominalising suffix -man in HC that derives nouns from verbs. The content of this section further investigates this claim. The Haitian form -man (from French -ment) is found in a number of nouns that can be related to verbs. For example, the noun rannman or randman 'yield' can be related to the verb rann 'to produce, to give off' (from Valdman et al 1981). The form -man thus appears to combine with verbs to form deverbal nouns. There are Haitian nouns containing -man that have no counterpart in French. Two examples are provided in (31). (31 )

HAITIAN

FRENCH

pledman 'discussion, quarrel, competition' kozman 'chat'

*plèdement/OK plaidoirie *causementlOK causette

As is shown in (32), there is an available verbal base for each of these nouns. (32)

pledman 'discussion, quarrel, competition' plede 'to argue (a case)' kozman 'chat' koze 'to chat' (from Valdman et al 1981)

As per the tests in section 10.1, the data in (31) and (32) suggest that nominalising -man is a potential affix of HC, deriving nouns from verbal bases. What about its productivity? The list of words ending in -man (with the exception of those involving adverbial -man, discussed in section 10.2.8) was copied from Freeman's inverse dictionary. Since Freeman's does not provide translations, words that are not listed in Valdman et aV s (1981) were eliminated from the list. Words ending in

280

ISSUES IN THE STUDY OF PIDGIN AND CREOLE LANGUAGES

nominal -man for which no base could be found were also excluded from the list. A sample of such nouns is provided in (33). (33) Noun tanperaman atachman dezagreman etoufman abiman chatiman alejman aranjman pikotman

'temperament' 'attachment' 'unpleasantness' 'suffocation' 'dress clothes' 'punishment' 'alleviation' 1.'agreement' 2. 'border, trimming' 'itch'

Base * ekoulman * abònman * anplasman * * * * *

anpresman trètman mekontantman apwentman antèman

* ransèyman

Base 'gonorrhea' * 'subscription' * 'building site' * (Fr: emplacement) 'haste' * 'treatment' * 'dissatisfaction' * 'salary, wages' * 'funeral, burial' *

'information' * (from Valdman et al 1981)

At this point, there remained 46 nouns in the list. Each noun was paired with a potential base. From this list, all the idiosyncrasies were eliminated. Idiosyncrasies are of various types. A first type involves lexical items ending in -man, where -man attaches to a base that is not verbal (recall that -man is claimed to derive nouns from verbs). The lexical items removed from the list on the basis of this criterion are listed in (34). In afrentan, -man is presumably attached to a nominal base that can also be used as an adjective. In fèman, -man is presumably attached to a nominal base. In balonnman and anbarasman, -man is presumably attached to attributive adjectives. (34)

afreman fèman balonnman anbarasman gonfleman egareman

'gluttony' afre 'machete' fè 'bloated stomach' balonnen 'financial difficulties' ambarase gonfle 'indigestion' 'foolishness' egare

'gluttonous' (N/A) 'iron' 'bloated' 'broke, without money' 'bloated' 'dumbfourded'

Even though there are no French words corresponding to the alleged derived Haitian words in the first two cases in (34), the latter must be analysed as idiosyncrasies. For example, in the case of fèman, it is difficult to construct the meaning 'machete' from the hypothesised internal structure of this word. Furthermore, the available base for the nouns in (34) is either nominal or adjectival. If the nouns in (34) were analysed as morphologically derived, they would be the only lexical items containing -man that would be derived from a non-verbal base. I thus conclude that the few lexical items in (34) do not argue in favour of the productivity of -man. Having removed these lexical items from the

THE EMERGENCE OF PRODUCTIVE MORPHOLOGY...

281

list, we are left only with lexical items made up of a potential verbal base and -man. In these cases, there is no or only partial semantic relationship between the potential verbal base and the presumed derived words. Such cases are listed in (35). For example, there is no semantic relationship between anpeche 'to prevent' and anpèchman 'financial difficulties'. In the case of deranje 'to disturb' and deranjman 'aching, intestinal desorder', there is only but a remote semantic relationship (as one can see deranjman as involving a 'disturbed digestive system'), but no direct one, where deranjman would refer to any kind of disturbance. (35)

anpèchman sakreman angajman deranjman bailsman rejetman

'financial difficulties' 'sacrament' 'pact with the devil' 'aching, intestinal disorder' 'mansion' bati 'vomiting' rejete

anpeche sakre angaje deranje

'to prevent' 'tonsured' 'to commit (oneself)' 'to disturb' 'to build' 'to renounce voodoo' (from Valdman et al 1981)

Since the meaning of the hypothesised derived words cannot be (fully) predicted from the meaning of their parts, the form -man in (35) cannot be argued to be productive. Another type of idiosyncrasy involves cases where the morphophonemics of the hypothesised derived words is not straightforward. Regular cases proceed as in (36). (36)

koz-man antann-man

< <

koze antann

On the basis of this model, we would expect -man to regularly attach to verb stems. As is shown in (37), however, the resulting structures are not always attested. (37)

*anbel-man *avil-man etc.

< <

anbeli avili

'to grow more beautiful' 'to degrade, to debase'

Instead, in such cases, the hypothesised derived words contain the sequence -isbetween the base and the suffix -man, as is shown in (38). (38)

anbelisman avilisman vomisman rafrechisman refwadisman

'beautification' 'debasement' 'vomiting' 'refreshment' 'chill'

anbeli avili vomi rajrechi refwadi

'to grow more beautiful' 'to degrade, to debase' 'to vomit' 'to refresh' 'to cool, cool'

282

ISSUES IN THE STUDY OF PIDGIN AND CREOLE LANGUAGES

toudisman

'dizziness'

toudi

avètisman sezisman

'warning' avèti 'shock (emotional)' sezi

'to stun, to daze, stunned, dazed' 'to warn, to notify' 'astonished, surprised'

The data show that the hypothesised derived nouns in (38) are not morphologically derivable in a straightforward way from the corresponding Haitian verbs. When we look at the nouns in (38) in light of the superstratum data, we find that these Haitian nouns correspond exactly to French nouns. These pairs of Haitian and French nouns are listed in (39). Note that the pronunciation of the Haitian and French words in (39) is almost identical; this fact may be obscured by the different orthographic conventions. (39)

HAITIAN

FRENCH

anbelisman avïlisman vomisman rafrechisman refwadisman toudisman avètisman sezisman

embellissement avilissement vomissement rafraichissement refroidissement étourdissement avertissement saisissement

Given the straightforward correspondances between the HC and the French nouns in (39), in addition to the fact that the HC nouns are not derivable in a straightforward way from their corresponding HC verbal bases, it is likely that the nouns in (39) have been integrally incorporated into HC from French. In any case, they do not argue for the productivity of -man. Consider the Haitian nouns and their potential verbal bases in (40). (40)

batisman 'construction (action)' bati 'to build' fini 'to finish' finisman 'end' remèsisman 'thank' remèsi/remèsye 'to thank'

The nouns in (40) cannot be claimed to be derived by means of the addition of -man to the verbal base. Such a rule would yield the unattested forms in (41). (41)

*bati-man *fini-man *remèsi-man

Unlike the nouns in (39), however, the nouns in (40) are not like the French corresponding ones, as is shown in (42).

THE EMERGENCE OF PRODUCTIVE MORPHOLOGY... (42)

HAITIAN

283

FRENCH

bailsman

'construction (action)' *batissement/OK bâtiment 'mansion' finisman 'end' *finissement/OK fin remèsisman 'giving of thanks' *remercissement/OK remerciement The Haitian nouns in (42) have thus not been integrally imported from French. As was pointed out in Orjala (1970: 154), these Haitian nouns appear to have been formed by analogy with the words in (39).11 Thus, they too have to be considered as idiosyncratic simplexes, and as such, they must be listed in the mental lexicon. When all the idiosyncrasies discussed so far have been removed from the list of 46 nouns, we are left with 23 nouns involving the nominalising form man. These naturally divide into two groups which I will discuss in turn. The first group contains 18 nouns that appear to be made up of a verbal base and man. The hypothesised deverbal nouns denote the action referred to by the verb or the result of this action. In all cases, the hypothesised derivation appears to be straightforward from a semantic, a morphological and a phonological point of view. These words are listed in (43). (43)

randman ankadreman santiman chajman soulaj(a)man ankourajman jijman chanjman gouvè(n)man kòmansman pansman konsantman derayman anmizman regleman debòdman tranbleman pèman

'yield' rann 'frame' ankadre 'feeling' santi 'cargo, load' chaje 'relief' soulaje 'encouragement' ankouraje 'judgement' jije 'change' chanje 'gouvernment' gouvènen 'beginning' kòmanse 'bandage' panse 'approval, consent' konsanti 'derailment' deraye 'entertainment' anmize 'settlement' regle 'overflowing' debòde 'trembling' tranble 'payment' peye

'to produce, to give off' 'to frame' 'to feel' 'to load, loaded' 'to relieve 'to encourage, encouraged' 'to judge' 'to change' 'to govern' 'to begin, begun, started' 'to bandage' 'to consent' 'to derail' 'to entertain' 'to settle, settled' 'to overflow' 'to tremble' 'to pay'

111 Orjala (1970: 154) considers the forms in (42) as Gallicisms. He notes: "In morphology, Gallicism seldom involves the adding of French inflection; but it does involve the employment of French derivational patterns in the creation by analogy of new words [...]."

284

ISSUES IN THE STUDY OF PIDGIN AND CREOLE LANGUAGES

antrénman antannman

'practice' 'understanding'

antrene antann

'to practice, to train' 'to get along'

Strikingly enough, however, all the pairs of nouns and verbs in (43) are also found in French (e.g. rendement 'productivity/yield', rendre 'to produce/to yield'; etc.). The Haitian data in (43) thus have nothing to contribute to the question of whether -man is native to HC. We are now left with the second group of nouns which contains the 5 items listed in (44). These appear to be derived in a straightforward way from a verbal base by means of -man. (44)

ankonbreman frapman vesman pledman kozman

'crowd, throng' 'knocking' 'vomit' 'discussion, quarrel' 'chat'

ankonbre frape vese plede koze

'to crowd, crowded' 'to hit, to bump, to strike' 'to vomit' 'to argue (a case)' 'to chat'

Unlike the pairs of deverbal nouns and verbs in (43), the pairs of deverbal nouns and verbs in (44) all have something that distinguish them from French. The first pair in (44) differs semantically from the corresponding French pair. In French, the verb encombrer means 'to congest, to encumber' and the deverbal noun encombrement means 'litter, congestion, traffic jam'. The second pair in (44) also differs from the corresponding French one. The French verb frapper 'to knock' cannot be nominalised, as *frappement is not attested. The third pair in (44) has a corresponding pair in French, but both the deverbal noun and the verb have a meaning that differs from the Haitian words. The French verb verser means 'to pour', and the French noun versement means 'pouring out'. The same observation carries over to the fourth pair in (44). The French deverbal nouns of the verb plaider 'to argue (a case)' are plaidoyer 'speech for the defense' or plaidoirie 'pleading' rather than *plaidement. Finally, the French word corresponding to Haitian kozman 'chat' is causette 'chat', rather than the nonattested *causement. So, the five deverbal nouns in (44) can be argued to have internal structure and to be derived by means of -man. Two additional deverbal nouns found in Sylvain (1936: 32) can be added to this short list. They figure in (45).12 (45)

degrès-man 'removing fat' defrip-man

degrese 'to remove fat, to lose weight' 'action of smoothing out' defripe 'to smooth out'

12 Sylvain (1936: 32) mentions that -man, as well as other potential affixal forms that she discusses, have very limited use.

THE EMERGENCE OF PRODUCTIVE MORPHOLOGY...

285

The conclusion to this exercise is that there are seven lexical items that appear to be derived by affixation of -man. Does this argue for the productivity of -man? Recall from section 10.2 that deverbal nouns can be created by means of morphological conversion and of suffixation with -ay. Morphological conversion was shown to apply to verbal bases that have only one argument, whereas affixation with -ay was shown to apply to verbal bases that have two arguments (where -ay absorbs one of the two). On the hypothesis that -man selects verbal bases, what are the properties of these bases in terms of their argument structure? The verbs that combine with -man are listed in (46) with their respective argument structure (based on Valdman et al. 1981). (46)

ankonbre koze frape degrese vese defripe plede

'to crowd' 'to chat' 'to hit' 'to lose weight' 'to vomit' 'to smooth out' 'to argue'

two arguments one argument two arguments one argument one argument two arguments one argument

It appears that the hypothesised suffix -man selects bases that have either one or two arguments. In contrast to the two other nominalising processes, it is thus not possible to characterise in a predictable way which verbal bases will be selected by -man. This casts doubt on the productivity of this potential affix. Another dimension that can be examined is the interaction between the hypothesised deverbal nouns ending in -man and those that occur as a result of conversion or of affixation with -ay. In Valdman et al. (1981), the verb koze 'to chat' in (46) can be nominalised by means of either conversion or affixation of -man, yielding koze 'chat' and kozman 'chat', respectively. According to Valdman et al. (1981), these two nouns are synonymous. There is yet another similar pair: tranblemanltranble 'trembling', both derived from the verb tranble 'to tremble', and also specified as being synonymous in Valdman et al (1981). In the literature on morphology, 'blocking' has been proposed to explain the non-existence of a complex form due to the existence of a synonymous competing one. Hence, in English, *gloriosity is not part of the lexicon because glory already exists. The fact that we find kozelkozman as synonyms in Haitian suggests that the two morphological processes, that is nominal conversion and -man affixation may have been competing at some point. Clearly, conversion has won the competition in terms of the number of tokens (see Lefebvre 1998a). The three other one-argument verbs in (46), vese, plede, and degrese, can only be nominalised by -man. Possibly, blocking prevents nominal conversion in these cases because of the existence of vesman, pledman and degrèsman.

286

ISSUES IN THE STUDY OF PIDGIN AND CREOLE LANGUAGES

In this case again, my conclusion is that it has not been proven that -man should be analysed as a productive affix of HC. 10.3.4 The hypothesised agentive suffix -ado DeGraff (2001: 60) claims that the form -adò(l) (phonologically derived from Spanish -dor) is a productive affix of Haitian. In support of his claim, he provides the three pairs of nouns and verbs in (47). These nouns are probably derived from the Haitian verbs on their right by means of the suffix -add. As is reflected by the translations in (47), the resulting deverbal nouns are agentive. (47)

babyadò vantadò kriyadò

'griper' 'braggart' 'chronic weeper'

babye vante kriye

'to whine' 'to brag' 'to weep' (from DeGraff 2001:60)

DeGraff (2001: 68) remarks that, if the suffix -ado "has its diachronic origins in cockfighting arenas near the Haitian-Dominican border (see e.g. Faine 1937: 45-46)", "it has now become fully integrated into HC lexicon and its scope includes concepts that are not at all related to cockfighting." The following examples illustrate his point. (48)

petado montado

'farter' 'liar'

pete manti

'to fart' 'to lie'

DeGraff (2001: 68) further notes the phonological difference between the form of the Haitian affix -add and that of its Spanish phonetic source -dor. The latter attaches to a verbal root with an intervening theme vowel, as in for example luchador 'fighter' and mentidor 'liar'. The corresponding verbs, luchar 'to fight' and mentir 'to lie', have distinct theme vowels, a and i. As HC babyadô from babye and mantadd from manti show, HC -add does not manifest theme-vowel distinctions: Spanish's most common theme vowel has been reanalysed as an integral part of the agentive suffix, leading to the creation of HC -add with its own combinatorial properties: -add attaches to appropriate verbal stems, independently of their verbal markers. (I thank Karlos Arregi for pointing out to me the HC-Spanish -adòl-dor differences.)

The contrast between the two forms argues for the specificity of the Haitian form with respect to its phonetic source. On the basis of these arguments, DeGraff (2001: 68) concludes that -add is a productive suffix of Haitian. As he puts it: "Notwithstanding its Spanish etymology, in the minds/brains of HC speakers -add is an "authentically Haitian" productive suffix." DeGraff does not provide any evidence supporting his claim that "in the minds/brains of HC speakers -add

THE EMERGENCE OF PRODUCTIVE MORPHOLOGY...

287

is an "authentically Haitian productive suffix"". This section further explores the properties and distribution of -add. A corpus of words ending in -ado was constituted in the following way. First, all the words ending in -adò in the inverse Haitian dictionary (Freeman 1988) were listed. The edition I have contains 18 such words. Second, an extensive search for Haitian words ending in -adò was carried out on the basis of the available literature. The following authors were consulted: d'Ans 1968; Faine 1937, 1974; Freeman 1988; Hilaire 1992, 1993; Orjala 1970; Peleman 1986; Pompilus 1976; Sylvain 1936; Tinelli 1970, 1981; Valdman 1996; Valdman et al 1981; Védrine 1992; Vernet et al 1976; Vilsaint 1992. The corpus that will serve as a database for the discussion below has the following features: The words that were listed in Freeman (1988) were retained, provided that they were also attested by at least one other author (because of lack of translation in Freeman). As a consequence of this procedure, one out of the eighteen words in Freeman had to be removed from the list: sangadò (unknown meaning). Words ending in -adò that were not part of Freeman's (1988) list, but that were mentioned by at least one author were also retained. This added only four lexical items to the initial list. The corpus so constituted is presented in (49). It contains 21 lexical items. The precise sources citing these lexical items will be identified later in the text. (49)

a. Words ending in -ado in Freeman (1988) that are also attested by at least one other author chichadò ponpadò bliyadò mizadò fouyadò woutchadò petadò kriyadò matadò paladò kouchadò vantadò plenyadò meladò sovadò babyadò karyadò b. Words ending in -adò(l) that were found in the literature but that were not listed in Freeman (1988) furadò montadò fumadòl kriyadò

I ran the list in (49) with an informant. This speaker has all the words in (49a) in his lexicon except for woutchadô and karyadô. Of the words in (49b), he has only montadò. Note that, in (49), there are two instantiations of kriyadô. As will be shown below, one is a borrowing from Spanish, while the other is a morphologically derived word. Even though the corpus in (49) is rather small, the 21 lexical items that it contains need to be organised in a principled way. Since the hypothesised Haitian suffix -ado derives its phonological representation from the Spanish form -dor, it is to be expected that, in the list in (49), there will be some lexical items that have been borrowed from Spanish. Consequently, the data in (49) can

288

ISSUES IN THE STUDY OF PIDGIN AND CREOLE LANGUAGES

first be divided into two groups: the lexical items that have been borrowed from Spanish, and those that have not. Criteria are needed in order to identify a given lexical item as a Spanish borrowing. First, the only lexical items that can be considered as having been borrowed from Spanish are those that cannot be argued to be derived from a Haitian base. In the list in (49), we are thus looking for lexical items that have no internal structure, that is, lexical items for which there are no available verbal bases in Haitian. Second, the lexical items that meet the first criterion have to have a corresponding Spanish word with similar phonological representation and shared meaning. In the corpus in (49), five lexical items meet the two requirements stated above. Two are attributive, the three others are agentive. They are listed in (50) with their corresponding Spanish word. (50)

Borrowings from Spanish a. Attributive deverbal noun kriyadd 'old, experienced' <Spanish (from DeGraff 2001) matado 'braggart' <Spanish (from Faine 1974: 463) b. Agentive deverbal nouns woutchadd 'fighter' <Spanish (from Faine 1937: 89) fumaddl 'smoker' <Spanish (from Faine 1937) kariadò 'the person who <Spanish puts two cocks face to face in preparation for a fight'

criador

'someone who raises' matador 'killer' luchador 'fighter' fumador

'smoker'

careador 'the person who takes care of the cock during the fight'
(from Freeman 1988; Faine 1937: 89) There is a potential verbal base for the noun kriyadò: the verb kriye. However, the semantics of this verb is different from that of kriyadd in (50a). While kriye means 'to weep' (see Valdman et al. 1981), kriyadd in (50a) means 'old, experienced'. So, this occurrence of kriyadd appears to have no internal structure in Haitian. Furthermore, there is a corresponding Spanish word with a related meaning: criador 'someone who raises'. Haitian kriyadd thus qualifies as a borrowing from Spanish. Matado is a borrowing from Spanish. Since there is no verb *mata- 'to kill' in Haitian, there is no Haitian base from which matado can be derived. As for woutchadò 'fighter', there is no available verbal base in Haitian, and there is a corresponding Spanish word luchador. It thus appears that woutchadd is a borrowing from Spanish luchador with some phonological adaptation. According to Valdman (1978: 141), fumaddl 'smoker' could be

THE EMERGENCE OF PRODUCTIVE MORPHOLOGY...

289

derived from the Haitian verbal base fimen 'to smoke'. If it were, we would expect the deverbal noun to have the unattested form *fimadòl. The form fumaddl finds a straightforward explanation if it is analysed as a borrowing from Spanish fumador 'smoker'. Finally, there is no available verbal base for the noun kariadô. But there is a corresponding Spanish noun careador with similar phonological representation and semantics. The word kariadô can thus be considered to be a borrowing from Spanish. There are thus five borrowed lexical items out of the 21 lexical items ending in -ado in (49). Since these lexical items are not derived in Haitian, they must be analysed as listed in the lexicon. There are 16 remaining words ending in -adò from the list in (49). One appears to be derived from an attributive word to which -add has been added: chichadô 'stingy person' < chick 'stingy' (from Valdman et al. 1981). Given the fact that this derivation is the only one of its kind, I will consider chichadô an idiosyncrasy listed in the lexicon. The 15 remaining words ending in -ado all appear to be derivable from a verbal base. The hypothesised verbal bases are all phonologically derived from a French verbal form. This excludes the possibility that the Haitian words so formed be analysed as borrowings from Spanish. In all cases the hypothesised derived word is a noun. The nouns so formed are all agentive. They are listed in (51). (51)

Haitian nouns ending in -ado and having a possible verbal base kouchadò 'sleepyhead' kouche 'to go to bed' 'to lie down' (from Valdman et al 1981) mizadò 'slowpoke' mize 'to dally, to dwadle' (from Faine 1937: 89, 1974: 272; Hall 1953: 41; Peleman 1986: 11; Sylvain 1936: 49; Valdman et al 1981) fouradò 'indiscrete' foure 'to put one's nose into...' (from Faine 1937: 89, 1974: 255;Hall 1953: 39; Peleman 1986; Sylvain 1936: 31; Valdman 1978: 141) babyadò 'griper' babye 'to whine; to grumble, to 'chronic complainer' complain' (from DeGraff 2001; Valdman et al 1981) yantado 'braggart' vante 'tobrag' (from DeGraff 2001; Hall 1953: 39, 41) kriyadò 'chronic weeper' kriye 'to weep' (from DeGraff 2001) petado 'farter' pete 'to fart' (from DeGraff 2001; Faine 1974: 340; Valdman et al. 1981; Vilsaint 1992: 103)

290

ISSUES IN THE STUDY OF PIDGIN AND CREOLE LANGUAGES

montadò 'liar' manti 'to lie' (from DeGraff 2001; Faine 1974: 463; Germain 1980: 86) sovadd 'runaway' sove 'to run away, to escape' (from Valdman et al. 1981) bliyadò 'forgetful person' bliye 'to forget' (from Valdman et al. 1981) fouyadò 'nosy person, fouye 'to dig, to dig up; to busybody' search, to frisk' (from Valdman et al. 1981) paladò 'talker' pale 'to talk' (from Faine 1937: 88, 1974: 64; Hall 1953: 39, 41; Peleman 1986; Sylvain 1936: 31; Valdman et al. 1981) plenyadò 'chronic complainer' plenyen 'to complain' (from Freeman 1988; Valdman et al 1981) meladò13

'someone who does mele not mind his/her own business' (from Freeman 1988; Peleman 1986) ponpadò 'refractory, stubborn' ponpe (from Freeman 1988; Peleman 1986)

'to interfere'

'to kick, to prance'

It would appear that there are 15 Haitian deverbal agentive nouns that can be claimed to be derived by means of the suffix -ado. How did the form -ado make its way into the Haitian lexicon? We can hypothesise that it got incorporated into Haitian in the same way as other cases observed in language contact situations did, that is, through bilingual speakers. Based on the scenario developed to explain similar cases (see e.g. Lefebvre 1984, and the references therein), I hypothesise the following scenario for Haitian -ado. (I disregard the phonological adaptations.) The bilingual speakers, Haitian-Spanish, who live in—or in the vicinity of—the Dominican Republic, incorporate Spanish words ending in -dor into their Haitian vocabulary.14 Having identified -dor as an agentive suffix, they start using it on verbal bases of 13 There is a Spanish word melado. It could thus be hypothesised that Haitian melado is in fact a borrowing from Spanish. This possibility is to be discarded, however, on the basis of the fact that Spanish melado has no semantics in common with Haitian melado. Indeed, Spanish melado means 'the colour of honey, juice from sugarcane or small honey cake'. 14 Speaking of the forms -adòi-adòl, Valdman (1978: 141) writes: "Il est fort probable que ces suffixes soient d'origine espagnole et soient parvenus au créole haïtien par voie de la République Dominicaine". [It is most probable that these suffixes are of Spanish origin and that they were incorporated into Haitian Creole from the Dominican Republic] Faine (1937: 49) makes the following observation: "En raison de l'émigration haïtienne vers les pays voisins, les mots espagnols continuent à s'insinuer sans cesse dans le créole." [Due to Haitian emigration to neighbouring countries, Spanish words keep being incorporated into the creole.]

THE EMERGENCE OF PRODUCTIVE MORPHOLOGY...

291

non-Spanish origin to form deverbal nouns such as those in (51). These new words may eventually become part of the vocabulary of the monolingual Haitian speakers. This scenario leaves us with two questions. Given the relatively small number of Haitian nouns in (51), do these nouns have internal structure for a significant number of monolingual Haitian speakers? That is: have monolingual Haitian speakers identified -add as an agentive deverbal suffix, or, are the nouns in (51) simply listed in the mental lexicon of these speakers? A second question is whether the hypothesised suffix -add is being used to form new words or whether the list of Haitian nouns ending in -add is pretty much frozen as is in (51). The remark by Sylvain (1936: 32) to the effect that -add (as well as other affixes that she discusses) is of very limited use, would suggest that the number of nouns formed with -add is not increasing. A clear answer to the two questions above requires tests that I leave for future research. It is worth mentioning the fact that -add is competing with the productive suffix -è deriving deverbal nouns having the same meaning as those formed with -add, that is, 'someone who VERBs'. It is possible that the productivity of -add is being blocked by the existence of -è. This possibility is supported by the fact that, out of the list in (51), there are four pairs of nouns ending in -add and in -è occurring with the same verbal bases, as is shown in (52). According to the cited dictionaries, these pairs of nouns are synonymous. (52)

petado 'farter' petè (from Faine 1937; Freeman 1988) montado 'liar' mantelmanto16 (from DeGraff 2001; (from Faine Faine 1974:463; Vernet et al Germain 1980: 86) 1996; Valdman vantadò 'braggart' vantè (from BFL 1989; Freeman 1988) ponpadò 'refractory' ponpadè (from Peleman 1986)

'farter'15 'liar' 1937: 89; Freeman 1988; 1976: 321; Valdman 1978, et al 1981; Vilsaint 1992: 155) 'braggart' 'refractory'

There is an additional pair of such words; in this case, however, the meaning of the two words differs in an interesting way; while sovadò refers to a 'runaway', sovè refers to a 'saver/saviour' (see Faine 1974; Freeman 1988). More competing forms of this type should be documented. They show that blocking works in creole languages as in any other language. 15 16

According to Freeman and Laguerre (1996),petèmeans 'sly or cunning person'. Freeman and Laguerre (1996) list manto as a variant of mantè 'liar'.

292

ISSUES IN THE STUDY OF PIDGIN AND CREOLE LANGUAGES

Pending additional research on this topic, I will conclude by assuming that it could be the case that -add has the status of a productive affix in Haitian Creole, but that the demonstration that it is the case has not been made as yet. 10.3.5 Summary The detailed investigation of the properties and distribution of the forms en-, -èt, -man and -adò casts doubt on their affixal status. On the one hand, in many cases, the words they are part of do not present the characteristics of words that have an internal structure. On the other hand, these forms are involved in a variety of idiosyncrasies including the category of the base, the semantics of the hypothesised base or of the hypothesised affix itself, and the properties of the words that they are part of. The forms en- and -èt were shown to not be productive at all. As for -man and -ado, our conclusion was that they have not been demonstrated as yet to be productive. An investigation of the remaining affixes shows that similar arguments can be made: -ab, -es and -te are unproductive, while the status of -syon is still under investigation (Denis in preparation). My conclusion is thus that, modulo a clear conclusion on -syon, the size of the inventory of the productive affixes in HC is limited to the affixes argued for in section 10.2.17 1 0 . 4 . The emergence of the morphological inventory of HC The inventory of the productive derivational affixes of HC in Table 10.1 comprises eleven affixes. Why is the morphological inventory of this particular size? Why does the inventory comprise this particular set of morphemes? How can the emergence of the HC morphological system be explained? This section addresses these questions within the framework of the relexification account of creole genesis. It summarises the scenario extensively discussed in Lefebvre (1998a: 303-33). In section 10.4.1, a comparison of the HC morphological inventory with those of its source languages shows that, although the forms of the HC affixes are derived from French, the size of the inventory as well as the properties of its affixes parallel those of its West African substratum languages. In section 10.4.2 the question of how this situation must have come about is addressed.

17 Valdman (1978) also considers the form -mann (<English man 'man') as a productive suffix of Haitian. [This form attaches to English bases (e.g. jazz-mann 'member of a jazz band'), as well as to an Haitian word that is phonologically derived from French (e.g. kaoutchoumann 'tire repair man'). Valdman (1978: 137) provides the following examples.

THE EMERGENCE OF PRODUCTIVE MORPHOLOGY...

293

10.4.1

The inventory of the productive derivational affixes of HC as compared with those of its contributing languages As is shown in Lefebvre (1998a), the Haitian derivational affixes all derive their phonological representations from French forms. The agentive suffix -è is derived from the phonetic matrix of the French agentive suffix -eur, as in travaill-eur 'worker'. 18 The attributive suffix -è is derived from the French attributive suffix -eur, as in farc-eur 'someone who makes jokes'. The verbalising suffix -e derives its phonological representation from the French suffix -er (pronounced [e]) as in fêt-er 'to celebrate'. The Haitian inversive/privative prefix de- is phonologically identical to the French inversive prefixed- occurring in verbs such as dé-placer 'to displace'. The Haitian prefix ti- is derived from Haitian pitit itself derived from the French form petit (pronounced [pəti]/[pti]/[ti]). The nominalising suffix -ay derives its phonological representation from the French suffix -age occurring in words such as chauff-age 'heating'. The adverbial suffix -man is derived from the French suffix -ment (pronounced [ma]): as in admirable-ment 'admirably'. Similarly, the Haitian suffixes -wa and -yen derive their phonological representations from the French suffixes -ois pronounced [-wa], as in village-ois 'from the village', and -ien as in parís-ien 'from Paris', respectively. Finally, the ordinal suffix -yèm derives its phonological representation from the corresponding French suffix -ième. When we compare the Haitian affixes with their French counterparts, however, a more complex picture arises. The productive Haitian affixes with their closest French corresponding forms are presented in Table 10.2 (along the lines of Brousseau 1994). The French forms that are not in parentheses are those hypothesised to have provided (i)

bos 'buggy' bos-mann 'coachman' gwòg 'grog' gwòg-mann 'someone who gets drunk' had 'frame' kad-mann 'someone who makes frames' djaz 'jazz orchestra' djaz-mann 'member of a jazz orchestra' kaoutchou 'tire' kaoutchou-mann 'someone who repairs tires' One Haitian informant consulted on this topic could think of only one example, (ii) radyo 'radio' radyo-mann 'someone who repairs radios' In my view, there is no evidence that the form -mann has aquired the status of a productive affix in Haitian. For example, the inverse dictionary (Freeman 1988) does not provide a single lexical entry ending in -mann. Furthermore, the only way this form could have entered the Haitian lexicon is through borrowing of English words containing it. Bilingual speakers Haitian-English are found among Haitians of the diaspora in some communities (e.g. American communities), but not all (e.g. French communities in France and in Canada), and certainly not in Haiti, where most inhabitants are monolingual in Haitian Creole. It is possible that, for the small portion of Haitians who are bilingual Haitian-English, the form -mann is becoming a productive affix, but, even for these speakers, the proof has not been made that it has on the basis of the frequency of words that contain this form.] 18 Albert Valdman points out that additional support of -è as a reflex of French -eur is provided by the existence of the variant -à as in mantè/mantò (e.g. Freeman and Laguerre 1996).

294

ISSUES IN THE STUDY OF PIDGIN AND CREOLE LANGUAGES

the phonetic matrices from which the phonological representations of the Haitian affixes were assigned; those in parentheses are forms that are synonymous (though they may differ in subcategorisation and morphophonemic properties). The data in Table 10.2 show that the derivational affixes of HC all have at least one phonetically similar corresponding affix in French. The most striking fact about the distribution in Table 10.2 is that in all cases, except for the agentive suffix, there are several French affixes corresponding to a single Haitian form (see BFL 1989: 18). For example, while French has several overt affixes converting verbs into nouns, Haitian has only one. Likewise, while French has several affixes designating a place of origin, Haitian has only two. Furthermore, it is notable that, as we saw in section 10.2.7, while Haitian has adjectival conversion, French does not. As is noted in Levin and Rappaport (1986), French derives adjectives from the past participial form of verbs rather than from the verbs themselves (e.g. La viande est cuite 'The meat is cooked'; la viande cuite 'the cooked meat'). Lefebvre (1998a: 314-318) provides a detailed comparison of the properties of the HC affixes and of their French corresponding forms showing discrepancies of various kinds between them: selectional properties, semantics of the derived words, relative productivity, position of the affix with respect to the base, etc. The conclusion of the comparison is that, while the phonological forms of the HC derivational affixes are derived from French forms, the properties of these affixes do not match those of the French ones. Finally, a most striking difference between the two inventories is their size. While the inventory of HC affixes in Table 10.2 exhausts the list of derivational affixes in this language, the inventory of French affixes in Table 10.2 is far from complete. For example, according to Dubois (1962), French has over 80 derivational suffixes (excluding Greek and Latin affixes involved in the vocabulary of the natural sciences, as well as all prefixes). Some authors claim an inventory of French productive affixes that is smaller than Dubois'. For example, Brousseau and Nikiema (2001: 289) estimate this inventory to comprise between 61 and 75 affixes, depending on the analysis (see also Thiele 1987). Even based upon these analyses, however, the inventory of French affixes far outnumbers that of HC. A comparison of the Hst of HC affixes in Table 10.2 with any list of French affixes immediately leads to the conclusion that the majority of the derivational affixes of French have no counterpart in HC.

THE EMERGENCE OF PRODUCTIVE MORPHOLOGY...

295

Table 10.2. The HC affixes and their closest French corresponding forms Agentive suffix base output Attributive suffix base output Verbalising suffix base output Inversive/privative prefix base output Diminutive prefix base output Nominalising suffix base output Conversion base output Adverbial suffix base output Place of origin/residence suffixes base output Ordinal suffix base output

HAITIAN -è V N -è N N -e N V deV V tiN N -ay V N

_

V N/A -man A Adv -wa/-yen N N -yèm Q A

FRENCH -eur V N -eur (-ard, -ier, -ien) N N -er, -é (-ir, -ifier, -iser) N/A V dé- (é-, in-, ir-) V/A V adjective petit [til (-et, -ot, -on) N/A N -age (-ion, -ment, -ance, -ure) V N - (-i, -é, -ert, -u) V N -ment A Adv -oisl-ien (-ais, -al, -and, -ain, -an) N N -ième Q A

(adapted from Lefebvre 1998a: 313-314) Lefebvre (1998a: 317) addresses the issue of whether this discrepancy could be attributable to the possibility that the inventory of French affixes was smaller at the time HC was formed, that is in 17th/18th century? A survey of the literature available on derivational affixes in 17th and 18th century French (see Julliand 1965; Nyrop 1936; Rey (ed.) 1992)19 and additional work by Tardif (ms) reveal that most of the affixes listed by Dubois were already in use at the time HC was formed. According to Tardifs survey, the major difference between classical and modern French lies in the productivity of specific affixes rather than in the size of the inventory of productive affixes. Thus, whatever the exact number of French productive affixes at the time HC was formed, we can safely conclude that they far outnumbered Haitian affixes. In the case of nominal conversion, 17th/18th century French still presented a few cases of deverbal nouns formed by this process. But, as is shown in Lefebvre (1998a: 314) this 19 Variation in the pronunciation of some derivational affixes between classical and modern French is discussed in Juneau (1972).

296

ISSUESINTHE STUDY OF PIDGIN AND CREOLE LANGUAGES

process is far more productive in HC than in the French of the time. This is because, while nominal conversion was very productive in Middle French, its productivity started to decline during the 16th and 17th centuries in favour of the use of overt affixes, as is stated by Haase (1975: 197-8) (our translation): The old language made frequent use of the infinitive with the sense of a noun, but this usage became rarer in the 17th century. ... in this period the nouns formed from infinitives were essentially the same as those still used in the language today: in addition to the very frequently used ones like le souvenir ['memory'], le lever ['rising'], le coucher ['bed' or 'setting'], le boire ['drink' or 'drinking'], le manger ['food'], etc., there are some rarer ones such as le marcher ['walking'], le pleurer ['crying'], le vivre ['living' or 'food'], ie vouloir ['will'], etc.

So, although French contributed the phonetic matrices of the Haitian affixes and, in most cases, their position with respect to the base they attach to, the comparison between the derivational affixes of Haitian and French leaves us with a number of questions: Why has Haitian developed this particular set of affixes? Why is the inventory so small compared with that of French? Why has Haitian developed adjectival conversion from verbs while French does it another way? Should the discrepancies between a significant number of Haitian and French derived words be considered as independent developments in Haitian? With these questions in mind, I now turn to the comparison of the inventory of Haitian affixes with that of one of its substratum language: Fongbe.20 Detailed descriptions of the productive derivational affixes of Fongbe can be found in Lefebvre (1998a: 312-8) and in Lefebvre and Brousseau (2002).21 Here, I only provide the comparison of the Haitian and Fongbe inventories based on these detailed descriptions. The comparative data are presented in Table 10.3. Note that the inventories in Table 10.3 are exhaustive for both languages. In the table, the term 'copy prefix' stands for the affix formed by partial reduplication of the base in Fongbe.22

20

Although there were several African languages spoken in Haiti in the late 17th century, the Gbe languages, among them Fongbe, have been identified as being predominant in Haiti at the time the creole was formed, that is at the end of the 17th century and beginning of the 18th century (see Singler 1996). See Lefebvre (1998a: 58-62) for a discussion of the typological features of the source languages of Haitian. 21 This description presents in a unified way information scattered in the literature on the morphology of Fongbe. Major sources cited are Brousseau (1990, 1993, 1994). 22 With Marantz (1982), we assume that reduplication is a kind of affixation and that the affix involved in reduplication is different from other affixes only in that it is phonologically underspecified, that is, it is only specified for a skeletal template. The segmental content of the template is determined by the phonological form of the base.

THE EMERGENCE OF PRODUCTIVE MORPHOLOGY... Table 10.3. The HC affixes and their Fongbe corresponding forms HAITIAN Agentive suffix -è base V output N Attributive suffix -è base N output N Verbalising suffix -e base N output V Inversive prefix debase V output V Diminutive affix tibase N output N Nominalising suffix -ay base V output N Conversion _ base V output N/A Adverbial suffix -man base A output Adv Place of origin/residence suffixes -wal-yen base N output N Ordinal suffix -yèm base Q output A

297

FONGBE -tó23

V N -no N N

_ màV/A V/A -Ví

N N copy prefix V N copy prefix V N/A

_ -tól-nù N N

-góó

Q A

(adapted from Lefebvre 1998a: 320-1) The Haitian and Fongbe derivational affixes in Table 10.3 differ in their phonological representation and they may also differ in their position with respect to the base they attach to. As we saw above, these properties of the Haitian affixes appear to have been largely provided by the superstratum language. In spite of these differences, however, there is a striking resemblance between the two inventories: in most cases, there is a one-to-one correspondence between the Haitian and Fongbe affixes. In both languages, there is one agentive, one attributive, one inversive, one diminutive and one ordinal affix. It is a remarkable fact that, in both Haitian and Fongbe, there are exactly two suffixes referring to a place of origin/residence. Furthermore, conversion in HC results in either nouns or adjectives; this corresponds to the output of the Fongbe derivations achieved by means of the copy prefix.24 The rather regular 23

Note that the suffixes -tó and -nó can also attach to synthetic compounds (see Lefebvre and Brousseau 2002: 230-231). [Partial reduplication of a verbal base converts verbal bases into nominal or adjectival ones. This process is observed in the Gbe languages; it is also widespread in West African languages and it is cited as a salient feature of the Kwa languages (see Westerman and Bryan 1970). While

24

298

ISSUES IN THE STUDY OF PIDGIN AND CREOLE LANGUAGES

correspondence between the Haitian/Fongbe affixes contrasts with the rather irregular correspondence between the Haitian/French affixes in Table 10.2. In spite of these similarities between the two inventories there are some differences between them. While the formation of deverbal nouns in HC involves two morphological processes, affixation of -ay and conversion, which are in complementary distribution, the formation of deverbal nouns in Fongbe involves phonological processes of the Caribbean creoles have generally been carried over into the creoles (see Brousseau in preparation for the Haitian Creole case), it appears that partial reduplication of the verbal base has not. This fact calls for an explanation. Pending further research on this topic, I can see two avenues for addressing this problem. The first one has to do with the complexification of the syllable structure in some cases such as Haitian (see Brousseau in preparation). The other one has to do with the fact that the patterns of partial reduplication are not at all uniform across West African languages, not even across Gbe languages (see Capo 1991). In contrast to partial reduplication, full reduplication of words is a widespread phenomenon in Caribbean creoles. As is the case in the substratum languages, full reduplication is used to convey intensification. Cases of this type of reduplication are reported in Jamaican Creole (see e.g. Cassidy 1957; DeCamp 1974), Sranan (see e.g. Sebba 1981), Saramaccan (see e.g. Bakker 1987), and Berbice Dutch (see Kouwenberg 1994). Haitian is no exception (see e.g. Hall 1953; Sylvain 1936; Valdman 1970), as is illustrated in (i). t' a tiretire. (i) M' I ANT IND.FUT shoot shoot T would have shot a lot.' (from Sylvain 1936: 83) In a few creoles, full reduplication of the base is used as a derivational process. Berbice Dutch uses full reduplication to form deverbal nouns as in (ii). (ii) koso koso 'the coughing' < koso 'to cough' (from Kouwenberg 1994: 249) Sranan also uses full reduplication of verbs to form deverbal nouns, as is shown in (iii). (iii) kamkamm 'comb' < kamm 'to comb' fumfum 'beating' < fumm 'to beat' (from Braun and Plag 2003) To my knowledge, these are the only two Caribbean creoles to form deverbal nouns by reduplication of the verbal base. Other creoles such as Haitian, for example, use morphological conversion to derive nouns from verbs or a nominalising affix, or both (see section 2). In Saramaccan, full reduplication of verbs yields participles and adjectives (see e.g. Bakker 1987; Muysken 1994b). The adjectives so formed can be either attributive or predicative, as is illustrated in (iv) and (v), respectively, (iv) dí fátufátu wómi the fat man 'the fat man' (from Bakker, Smith and Veenstra 1995: 171) (v) dí wómi dé nákináki the man be. at beaten 'the man has been beaten (is in a beaten state)' (from Bakker, Smith and Veenstra 1995: 171) As for Papiamentu, it presents some cases of reduplication. According to Kouwenberg and Murray (1994) based on Maurer (1988), and Dijkhoff (1993), however, this process does not derive nouns nor adjectives from verbs in a regular and productive way. The reason why some Caribbean creoles manifest reduplication as a productive morphological process and others do not is a question for future research. This question is all the more intriguing since Haitian, Papiamentu, Sranan and Saramaccan all share the same substratum languages.]

THE EMERGENCE OF PRODUCTIVE MORPHOLOGY...

299

only one affix: the copy prefix. This mismatch is explained in the following way in Lefebvre (1998a, section 10.4). While reduplication is widely used among Kwa languages to derive adjectives from verbs (see Westerman and Bryan 1970), its use to derive nouns from verbs is restricted to a subset of the Kwa languages. Fongbe is part of this subset. Other Kwa languages such as Twi, Ewe, Yoruba and Nupe use another affix. This is illustrated in (53). (53)

a. b. c. d.

a-dow o-gblo -da i-bi

'hoeing' 'breadth' 'creature' 'wickedness'

<dow 'hoe' TWI
So, in this case, HC would reflect the morphological division of a subset of its substratum languages that excludes some Gbe languages such as Fongbe. A second mismatch concerns the verbalising affix -e and the adverbial affix -man, which have no counterparts in Fongbe. In Lefebvre (1998a: 323-333) these mismatches are considered as cases of innovations from within the creole, made possible by the availability of French forms. Derived words in Haitian and Fongbe can also be contrasted from the point of view of their semantics. As is shown in Lefebvre (1998a: 312-318; 403-408), there is quite a large number of Haitian derived words which do not correspond to derived words in French. Lefebvre (1998a: 318-323) further shows that, in most such cases, there is a Fongbe counterpart to the Haitian word. For example, while there is no French derived word corresponding to Haitian eskandal-è 'loud, rowdy', there is a corresponding Fongbe word derived by means of the attributive suffix -no: zîgdí-nò 'loud, rowdy'. Likewise, while cases of nominal conversion in Haitian do not all have a corresponding lexical item so derived in 17th/18th century French (see Lefebvre 1998a: 312-318), all the Haitian deverbal nouns have a corresponding Fongbe deverbal noun derived by means of the copy prefix, showing that nominal conversion in Haitian and the 'copy prefix' in Fongbe have a similar function. Likewise, while Haitian nouns derived by means of the nominalising suffix -ay often do not have a French corresponding noun derived in this way, in most cases, there is a Fongbe deverbal noun derived by means of the copy prefix that corresponds to the Haitian noun with -ay (see Lefebvre 1998a for a detailed discussion of these facts and for data illustrating each case). These facts strongly support the claim that the concatenation of particular affixes with particular bases in Haitian follows the pattern of the substratum languages, and that the mismatches between Haitian and French can, to a large extent, be explained by the claim that

300

ISSUES IN THE STUDY OF PIDGIN AND CREOLE LANGUAGES

the creators of Haitian used the properties of their own lexicons in concatenating affixes and bases in the language that they were creating. Finally, although the correspondences between the two inventories in Table 10.3 are not perfect, the striking fact about them is that they are of comparable size. Indeed, the list of productive affixes in Table 10.3 exhausts the number of productive affixes in both Haitian and Fongbe. This strongly suggests that the size of the HC inventory of productive derivational affixes has, to a large extent, been determined by that of the substratum lexicons. From the discussion in this section, it appears that, while French has contributed the forms, and in most cases the position, of the HC productive derivational affixes, the substratum languages have, to a large extent, contributed the semantics of these Haitian affixes, the principles which govern their concatenation with their bases and the extent of the inventory. How can such a division of properties be accounted for? 10.4.2

The relexification account of the emergence of the HC morphological inventory Relexification is a cognitive process that consists in the relabelling of a lexical entry of a given language on the basis of a phonetic string taken from another language.25 In Lefebvre (1998a and the references therein), it is extensively argued that the bulk of the HC lexicon has been created by this process (see also chapter 3). Following this theory, the creators of HC, speakers of West African languages, relabelled the lexical entries of their respective lexicons on the basis of French phonetic matrices. The relexified lexicons constituted the incipient creole. This theory of creole genesis, supported by a large body of HC data, explains why the creole lexicon manifests the properties of its contributing languages in the way it does: while the phonological representations of the creole lexical entries are derived from phonetic matrices in its superstratum language, the bulk of its other properties are derived from its substratum languages. Since derivational affixes are lexical categories that are listed in the lexicon, they are assumed to undergo relexification in the same way as other lexical categories do. In Lefebvre (1998a: 323-333), it is thus argued that the creators of HC have relabelled the affixes of their own lexicons on the basis of French phonetic matrices. In this scenario, the creators of Haitian would have identified phonetic matrices of French which had an appropriate meaning to relabel the derivational affixes' lexical entries of their own lexicons on the basis of pairs of French words such as faire 'to do' and dé-faire 'to undo', or travaill25

From the point of view of second language acquisition, the process of relexification thus constitutes an extreme case of interference.

THE EMERGENCE OF PRODUCTIVE MORPHOLOGY...

301

er 'to work' and travaill-eur 'worker'. Hence, a Fongbe speaker relexifying his/her lexicon on the basis of data from French would have identified French dé- as sharing with mà- the meaning 'inversive' and (s)he would have assigned the Haitian lexical entry corresponding to mà- the phonological form dé-. Similarly, the agentive affix -to would have been relabelled as -è on the basis of the French agentive affix -eur, and so on. This scenario accounts for the fact that there is a one-to-one correspondence between the HC and Fongbe agentive, attributive, inversive, diminutive, ordinal and place of origin affixes. Since by definition, the copy prefix does not have an independent phonological representation, it cannot be claimed to have been relexified. The word formation rule involving the copy prefix in Fongbe, however, systematically corresponds to cases of morphological conversion in HC. In both cases the morphological process produces nouns and adjectives. The fact that morphological conversion in HC produces adjectives as well as nouns thus finds a straightforward explanation in the substratum languages. Finally, the relexification account explains in a simple way why the extra French affixes, a subset of which are listed in Table 10.2, did not make their way into HC: there were no such lexical entries in the original lexicon to be relexified. (The two Haitian innovations -man and -e and the presence of the additional suffix -ay were discussed in section 10.4.1.) As has been pointed out in Lefebvre and Lumsden (1989), the relexification account of creole genesis constitutes a further development of the second language acquisition theory of creole genesis (e.g. Alleyne 1980; Andersen 1980; Schumann 1978; Thomason and Kaufman 1991; Valdman 1980; etc.). Relexification is seen as a tool for acquiring a second language in a context where there is reduced access to the second language. The role of relexification in the formation of creoles was hypothesised to be central on the basis of the fact that, in situations where creoles emerge, there is reduced access to the superstratum language. In the case of HC, the hypothesis was supported by both linguistic and historical data. On the one hand, Lefebvre (1998a, 1999a, and the references therein) extensively shows that the bulk of the HC vocabulary can be argued to have been created by means of relexification. On the other hand, Singler (1996) establishes that HC was formed between 1680 and 1740; he shows that the beginning of this period was marked by a shift from a tobacco and cotton to a sugar economy, and that this shift provoked a radical change in the configuration of the early Haitian population: "the large land owners drove out the small ones, the practice of importing engagés ceased, and the number of slaves exploded" (Singler 1996: 193). As is observed by Singler (1996), this shift in the composition of the early Haitian population had the effect of modifying the slave population's exposure to French. Indeed, under the first

302

ISSUES IN THE STUDY OF PIDGIN AND CREOLE LANGUAGES

type of economy, the slave population was in frequent contact with French through the landowners, engagés and people of mixed race. In the shift to a sugar economy, the engagés disappeared from the Haitian population and the enslaved population increased dramatically, greatly reducing day-to-day contacts between French speakers and the bulk of the Haitian African population. In light of this historical information, the reason why only a small portion of French morphology made its way into HC finds further explanation. It is hypothesised that, because they did not have enough access to French, the creators of HC generally did not have access to the internal structure of French words. By hypothesis, in the French material that they were presented with, they were looking for phonetic matrices that corresponded to the productive morphemes that they had in their own lexicons. Thus, on the basis of a few pairs of words, they could identify morphemes that corresponded to the productive ones in their respective lexicons. The great majority of the French affixes had no corresponding ones in the West African lexicons and thus they were not identified as such. The French words containing affixes that had no equivalents in the substratum languages, were incorporated wholesale in the new lexicon. A final point needs to be considered for this scenario to be complete. Recall from section 10.3 that there are morphological forms that could be identified as potentially native to Haitian but that turned out to be unproductive. How do these data fit into the relexification account summarised above? In section 10.4.1, we saw that the superstratum may offer several forms corresponding to only one or two in the substratum languages. For example, while French has various inversive prefixes such as dé-, é-, in-, ir-, there is only one such prefix in Fongbe, mà- (see Table 10.3). Likewise, while French has several nominalising affixes: -age, -ion, -ment, -ance, -ure, Fongbe has only one, used to nominalise all the verbs, and Ewe has two (see the above discussion and the example in (53)). In the same fashion, Fongbe has two suffixes encoding the place of origin/residence, but French has many more: -ois, -ien, -ais, -al, -and, -ain, -an, and so on and so forth. Thus, at the time relexification took place, there were several available forms in the superstratum language to relexify each substratum form. The comparisons of Haitian and French forms in Table 10.2 and of Haitian and Fongbe ones in Table 10.3 show, however, that, out of the available French forms, the creators of Haitian selected the number of forms that they already had in their own inventory. I explain these facts in the following way: Before the Creole's inventory attained its present form, there was probably an unstable period where there were more forms that were potentially native to Haitian than those that can be argued to be productive in the contemporary language. The existence of forms identified in section 10.3 as potentially native to HC on the basis of one of the tests in section

THE EMERGENCE OF PRODUCTIVE MORPHOLOGY...

303

10.1, but that turned out to be unproductive on closer scrutiny, supports the hypothesis of an early unstable period in the development of the morphology of the creole. This period would be characterised by the presence, in the creole, of several forms competing for the same lexical (morphological) entry. For example, in section 10.3.3 we saw that the form -man was competing with -ay and with morphological conversion. The few words ending in -man that appear to have internal structure in (44) and (45) are possibly leftovers from this period. Furthermore, pairs such as those in (54) probably also reflect the intermediary unstable period hypothesised here. In (54), the pairs of deverbal nouns have two different derivations but exactly the same meaning (they are considered as synonyms in Valdman et al 1981). The first deverbal nouns in (a) and (b) are derived by conversion yielding nouns that have the same form as the corresponding verbs; the second deverbal nouns in (a) and (b) are apparently derived by affixation of -man. (54)

a. b.

koze trenble

'chat' 'trambling'

kozman trenbleman

'chat' 'trambling'

The hypothesis according to which the apparent derivations in -man in (54) would constitute leftovers from the hypothesised unstable period under discussion is congruent with the fact that blocking precludes the derivation of kozman and of tranbleman because of the existence of the deverbal nouns koze and tremble produced by morphological conversion. 10.4.3 Summary The comparison of the derivational affixes of Haitian, French and Fongbe presented in this section shows that, while the Haitian derivational affixes derive their phonological representations from French phonetic matrices, they derive most of their other properties from their substratum languages. The principled division of properties of the HC affixes between its source languages was shown to follow from the relexification account of creole genesis, as formulated and argued for in Lefebvre (1998a, and the references therein). 1 0 . 5 . Conclusion: Issues on the morphology of creole languages in light of the HC data There is a long standing belief that pidgin and creole languages have poor mor phology, if any at all. For example, d'Ans (1968: 26) writes: "One of the salient characteristics of the creole is precisely to not present processes of derivation, and even to logically not be able to present such processes" (our translation). Hall (1953: 34) remarks that "the three processes of derivation (prefixation, suffixation, and compounding) are for the most part rare and are found but

304

ISSUES IN THE STUDY OF PIDGIN AND CREOLE LANGUAGES

scatteringly in creole, and to a large extent only in learned words borrowed from standard French." While it seems to be the case that pidgin and creole languages generally manifest little inflectional morphology (e.g. Mufwene 1986, 1990, 1991), it is not the case that these languages lack morphology all together. On the one hand, compounding appears to be quite productive in these languages (for Sranan, e.g. Adamson and Smith 1995; Braun and Plag 2003; for Saramaccan, e.g. Bakker, Smith and Veenstra 1995; for Papiamentu, e.g. Dijkhoff 1993; for HC, e.g. Brousseau 1989; Valdman 1978; for Berbice Dutch, e.g. Kouwenberg 1994; for Fa d'Ambu, e.g. Post 1995; for French-based creoles of the Indian Ocean, Papen 1978; etc.). On the other hand, creoles do exhibit derivational morphology to varying degrees. For example, as was shown in this chapter, there are at least eleven affixes that can be argued to be productive in HC. Even in more recent literature, it has been claimed that creole languages are morphologically poor. For example, Seuren and Wekker (1986: 62) have proposed the Semantic Transparency Hypothesis which "stipulates a strict oneto-one correlation between semantic elements and structural surface units (morphemes)"; in their view, semantic transparency "limits learning and computing efforts to a minimum" (p.68). [In chapter 7, it was shown that this hypothesis is not borne out by data drawn from the lexical, syntactic, or interpretive components of Haitian grammar. In this section, I question the relevance of this hypothesis on the basis of data drawn from the morphological component of the Haitian grammar.] According to Seuren and Wekker, the Semantic Transparency Hypothesis predicts that, among other things, creole languages will lack morphology, where lack of morphology is considered as a means of "maximising semantic transparency" (p.62). The HC data constitute a direct counterexample to this claim in manifesting a significant number of productive derivational affixes. Indeed, the size of the HC inventory is impressive in view of the fact that some languages, not known to be creoles e.g. Vietnamese, do not even have one derivational affix. The HC data constitute a counterexample to the Semantic Transparency Hypothesis from yet another point of view. In the HC inventory, there is not necessarily a "one-to-one correlation between semantic elements and structural surface units." For example, the agentive and the attributive suffixes are homophonous. The form de- is either inversive or privative. Furthermore, Haitian has two suffixes used to form nouns that refer to a place of origin, and the principles governing the choice between the two affixes -wa and -yen —either semantic or phonological—, if any, are not known. The Semantic Transparency Hypothesis would predict that there should be only one (if at all) affix deriving nouns that refer to a place of origin in HC. The Semantic Transparency

THE EMERGENCE OF PRODUCTIVE MORPHOLOGY...

305

Hypothesis also predicts a lack of semantic extensions and idiosyncrasies in the lexicon of creoles. As we saw in section 10.3, and as will be further discussed below, semantic extensions and idiosyncrasies abound in HC. In contrast to Seuren and Wekker, McWhorter (1998a) allows for the presence of derivational affixes in creole languages. He claims, however, that "in languages known as creoles, derivation is generally semantically transparent" (1998a: 797). In the theoretical approach adopted here, productive morphology is, by definition, semantically transparent since affixes have categorial features, meaning, selectional properties, etc. In this view, the productive morphology of a creole language cannot be claimed to be more transparent than the productive morphology of a non-creole language. In the context of the Bickertonian view that creole languages would reflect Universal Grammar in its unmarked form (e.g. Bickerton 1984), it has been proposed that the size of a Creole's morphological inventory is determined by universal principles (e.g. Mühlhäusler 1980: 36). The relexification account of creole genesis presented in section 10.4 rather shows that the morphological inventory of a given creole is, to a large extent, determined by that of its substratum languages, provided that suitable forms can be found in its superstratum language. Furthermore, the comparison of the morphological inventory of HC with those of its contributing languages shows that if HC has fewer productive affixes than French, its lexifier language, it has more productive affixes than Fongbe, one of its substratum languages (see Table 10.3). The claim by Thomason (2001: 168), according to which pidgins and creoles "have very limited morphological resources compared with those of the lexifier and other input languages", is thus not borne out by the HC data. McWhorter (1998a) holds that semantic drift creates endless lexicalisations over time. According to him, since languages that are known to be creoles are relatively young, there has not been time for idiosyncratic lexicalisations to be formed in these languages. McWhorter's view thus predicts a lack of semantic extensions and idiosyncrasies in the lexicon of creole languages. This prediction is far from being borne out by HC data. Indeed the Haitian lexicon is full of idiosyncrasies of all kinds, as was already shown in section 10.3. The following paragraphs add further examples. In section 10.2.4, we saw that de- productively derives new verbs that have either an inversive or a privative meaning with respect to the base it attaches to. There are several Haitian words beginning with the form de-, which at first glance, may look like the prefix de-. Closer examination of these words, however, shows that in these cases, de- cannot be associated with the productive affix de bt it inversive or privative. For example, consider the word debaba 'to mow (a lawn), to weed' (from Valdman et al. 1981). Not only is there no verb *baba, but

306

ISSUES IN THE STUDY OF PIDGIN AND CREOLE LANGUAGES

there is no noun *baba with a related meaning. The lexical entry baba in Valdman et al (1981) is glossed as 'idiot'. It thus appears that baba and debaba are not morphologically related. (See also DeGraff 2001 for discussion of this pair.) Debaba must thus be listed in the lexicon. The Haitian verb demele constitutes another idiosyncrasy. According to Valdman et ál. (1981), the verb demele means 'to untangle', 'to get along (manage)' and 'to shuffle (cards)'. The action of shuffling cards has the effect of mixing the cards rather than that of untangling them. This meaning is thus closer to that of mele 'to mix, to blend'. This is a rare type of idiosyncrasy. Another set of data concerning de- is presented in (55). In this case, de- occurs with a verbal base, derived from a noun. For example, the noun plim 'feather, body hair, fur' is converted into a verb by means of the verbalising suffix -e yielding plimen 'to pluck'. This verb may occur with de-, yielding de-plimen which also means 'to pluck'. Thus, with or without de-, the verb has the same privative interpretation. In Valdman et al (1981), the pairs of verbs in (55) are given the same definition. (55)

degrennen 'to shell' grennen 'to shell' < grenn dekale 'to chip off, kale 'to shell, < kal to peel off' to peel' deplimen 'to pluck' plimen 'to pluck' < plim derefize devire

'to refuse' 'to turn'

refize vire

'seed' 'covering(bark, skin, shell)' 'feather, body hair, fur' 'to refuse' < refi(z) 'refusal' 'to turn' < vire 'turn'

The data in (56) present a similar pattern of pairs of verbs. In this case, however, there are no nouns involved in the derivation of the verbs. (56)

dechire, chire dechifonnen, chifonnen degengole, gengole

'to tear, to rip' 'to wrinkle, to rumple' 'to rush down' (from Valdman et al. 1981)

Some pairs of adjectives also follow this pattern, as is shown in (57). (57)

demefyan, mefyan

'mistrustful'

(from Valdman et al. 1981)

In (55), (56) and (57), the form de- is simply semantically vacuous.26 The existence of pairs of synonyms, such as those in (55), (56) and (57), is certainly 26

DeGraff (2001) considers as emphatic the verb derefize as opposed refize in (47). He also considers as emphatic demegri, as opposed to megri, both meaning 'to lose weight'. To my knowledge, no other author provides these emphatic interpretations for words containing the prefix de-. Furthermore, an emphatic interpretation of the other verbs in (47) (e.g. dekale 'to chip off, to peel off', deplime 'to pluck', etc.), is hard to imagine. Albert Valdman remarks,

THE EMERGENCE OF PRODUCTIVE MORPHOLOGY...

307

not compatible with the view that there are no idiosyncrasies in a Creole's lexicon. Furthermore, from the learner's point of view, the above data cannot be considered as being semantically transparent. Another type of idiosyncrasy involving the form de- consists in the semantic extensions found with verbs containing the productive prefix de-. Examples are given in (58). (58)

degwosi dechouke

'to lose weight, to smooth out, to rough-hew' 'to get someone fired, to remove from office, to uproot, to plow' dechaje 'to unload, to ejaculate' degagannen 'to cut the throat of, to overcharge' degonfle 'to deflate, to scatter, to make room for' degrade 'to demote, to damage, to take away the supernatural power of a dead person' degrennen 'to shell, to break a bill, to knock out, to thin out (forest)' dekoupe 'to eat copiously, to cut up, to interrupt' dekreta 'to cut of the comb of a rooster, to mistreat' demanbre 'to batter (a person), to ravage, to weaken, voodoo dwelling left by the ancestors' detripe 'to gut, to defame' (from Valdman et al 1981)

The semantic extensions of the kind found in (58) are idiosyncratic and they must be listed in the lexicon (see also DeGraff 2001).27 In section 10.2.5, we saw that the prefix ti- productively derives denominal nouns having the meaning 'a young NOUN' or 'a small NOUN'. There are semantic extensions of words containing the form ti-. These words have to be listed as such in the lexicon because, in contrast to words that have internal structure, in these cases, the whole is not predictable from its parts. These semantic extensions are manifested in euphemisms such as those in (59), or in expressions such as those in (60).

however, that "the emphatic connotation is suggested by the artful manner in which, in his novel Dezafi, the writer Franketienne piles up derivatives with de-" This remark suggests that the emphatic connotation associated with de- could come from the succession of words beginning with de-. The question here is whether the emphatic interpretation claimed to be associated with de- is part of the meaning of de-, or a stylistic effect. 27 The discussion of the source of these idiosyncrasies goes far beyond the scope of this chapter. This topic, however, is a goldmine for future debates on creole genesis.

308

ISSUES IN THE STUDY OF PIDGIN AND CREOLE LANGUAGES

(59)

ti-bezwen ti-devan ti-wòz ti-gigit ti-pijon

(60)

ti-bourik ti-fi ti-gran moun ti-grann ti-malis ti-nouris ti-tèt

(bezwen 'need') (devan 'front') (wòz 'rose') (gigit 'penis') (pijon 'pigeon')

'genitals (of a small boy)' 'genitals' 'menstrual period' 'penis' 'penis' (from Valdman 1996; Valdman et al 1981)

(bourik 'donkey') (fi 'woman') (gran moun 'adult, old person') (grann 'grandmother') (malis 'folk character known for his cleverness') (nouris 'wet nurse, nursing mother') (tèt 'head') (from Valdman

'rude person' 'virgin' 'precocious child' 'precocious little girl' 'canny person' 'new mother' 'birdbrain' 1996; Valdman et al 1981)

Morphological conversion in section 10.2.7 was shown to derive nouns from verbs in a very productive way. Similar cases, whose properties are not predictable, were considered to be relisted in the mental lexicon. Likewise, the data discussed in section 10.3 constitute idiosyncrasies that need to be listed in the mental lexicon. In addition to the data in section 10.3, the above data demonstrate in an unequivocal way the existence of various types of idiosyncrasies in the HC lexicon. They show that idiosyncrasies are no 'exceptions' in this lexicon. McWhorter's (1998a) prediction that there has not been time for idiosyncratic lexicalisations to be formed in creole languages is most certainly not borne out by the HC data. This conclusion is even more striking considering the fact that HC has only had 200 years of independent existence.28 This conclusion is congruent with that of Braun and Plag (2003) who found good evidence for idiosyncrasies in Sranan when it was only one hundred years old.

28

Singler (1996) establishes the creation of HC between 1680 and 1740. According to Hilaire (1993), African languages were spoken in Haiti up until the beginning of the nineteenth century. At the time of the Revolution, in 1791, the by-then well established creole took over. (For further discussion, see Singler 1996 and Lefebvre 1998a: 52-58.)

REFERENCES Abney, S. 1987. "The English noun phrase in its sentential aspect." Unpublished Ph.D. dissertation, MIT, Cambridge, MA. Adam, L. 1883. Les idiomes négro-aryen et maléo-aryen. Paris [No indication of publisher]. Adamson, L. and Smith, N. 1995. "Sranan." In J. Arends, P.C. Muysken and N. Smith (eds), 219-232. Adone, D. 1994. The Acquisition of Mauritian Creole [Language Acquisition & Language Disorders 9]. Amsterdam/Philadelphia: John Benjamins Publishing Company. Adone, D. and Plag, I. (eds). 1994. Creolization and Language Change. Tübingen: Niemeyer. Agbidinoukoun, C.C. 1991. "Analyse contrastive des syntagmes nominaux du fongbe et du français." Unpublished Ph.D. dissertation, Université Paris III (Sorbonne Nouvelle), Paris. Allen, M.R. 1978. "Morphological investigations." Unpublished Ph.D. dissertation, University of Connecticut, Storrs, CT. Alleyne, M.C. 1966. "La nature du changement phonétique à la lumière du créole français d'Haïti." Revue de linguistique romane 30: 279-303. Alleyne, M.C. 1971. "Acculturation and the cultural matrix of creolization." In D. Hymes (ed.), 169-187. Alleyne, M.C. 1980. Comparative Afro-American: An Historical-Comparative Study of English-Based Afro-American Dialects of the New World. Ann Arbor, MI: Karoma. Alleyne, M.C. 2000. "Opposite processes in "creolization"." In I. NeumannHolzschuh and E.W. Schneider (eds), 125-133. Amberber, M. 1996. "The transitivity of verbs of saying revisited." In Proceedings of the Fourteenth West Coast Conference on Formal Linguistics, J. Camacho, L. Choueiri and M. Watanabe (eds), 1-15. Stanford, CA: Center Study Language & Information. Andersen, R.W. 1980. "Creolization as the acquisition of a second language as a first language." In A. Valdman and A.R. Highfield (eds), 273-295. Andersen, R.W. 1983. "A language acquisition interpretation of pidginisation and creolisation." In R.W. Andersen (ed.), 1-59. Andersen, R.W. (ed.). 1983. Pidginisation and Creolisation as Language Acquisition. Rowley, MA: Newbury House. Anonymous. 1983. Éléments de recherche sur la langue fon. Cotonou [No indication of publisher]. Ans, A.-M. d'. 1968. Le créole français d'Haïti : Étude des unités d'articulation, d'expansion et de communication. The Hague: Mouton de Gruyter.

310

ISSUES IN THE STUDY OF PIDGIN AND CREOLE LANGUAGES

Aoun, J. 1992. "The search for facts." Textes des séances plénières, 3e circulaire, XVe Congrès international des linguistes, 77-88. Quebec: Université Laval. Arends, J. 1986. "Genesis and development of the equative copula in Sranan." In P.C. Muysken and N. Smith (eds), 103-127. Arends, J. 1989. "Syntactic developments in Sranan." Unpublished Ph.D. dissertation, Catholic University, Nijmegen. Arends, J. 1995. "Demographic factors in the formation of Sranan." In J. Arends (ed.), 233-277. Arends, J. 2002. "The historical study of creoles and the future of creole studies." In G.G. Gilbert (ed.), 49-68. Arends, J. (ed.). 1995. The Early Stages of Creolization [Creole Language Library 13]. Amsterdam/Philadelphia: John Benjamins Publishing Company. Arends, J., Muysken, P. and Smith, N. (eds). 1995. Pidgins and Creoles. An Introduction [Creole Language Library 15]. Amsterdam/Philadelphia: John Benjamins Publishing Company. Aronoff, M. 1976. Word Formation in Generative Grammar. Cambridge, MA: The MIT Press. Aronoff, M. 1980. "The relevance of productivity in a synchronic description of word formation." In Historical Morphology, J. Fisiak (ed.), 71-82. The Hague/Paris/New York: Mouton de Gruyter. Aronoff, M. 1983. "Potential words, actual words, productivity and frequency." In Proceedings of the XIII International Congress of Linguists, August 29-September 4, 1982, Tokyo, S. Hattori and K. Inoue (eds), 141-148. Tokyo: Permanent International Committee on Linguistics. Aronoff, M. and Schvaneveldt, R. 1978. "Testing morphological productivity." Annals of the New York Academy of Science: Papers in Anthropology and Linguistics 318:106-114. Avolonto, A. 1992. "De l'étude sémantico-syntaxique des marqueurs préverbaux à la structure de la phrase en fòngbè." Unpublished MA thesis, Université du Québec à Montréal, Montreal. Awoyale, Y. 1986. "Reflexivization in Kwa languages." In Current Approaches to African Linguistics [Publications in African Languages and Linguistics 3], G.J. Dimmendaal (ed.), 1-14. Dordrecht: Foris Publications. Baker, M.C. 1991. "On the relation of serialisation to verb extensions." In C. Lefebvre (ed.), 79-102. Baker, M.C. and Stewart, O.T. 1996. "Unaccusativity and the adjective/verb distinction: Edo evidence. " Paper presented at NELS 27, McGill University, Montreal. Baker, P. 1990. "Off target?" Journal of Pidgin and Creole Languages 5 (1): 107-119. Baker, P. 1993. "Assessing the African contribution to French-based creoles." In S.S. Mufwene (ed.), 123-155. Baker, P. 2000. "Theories of creolization and the degree and nature of restructuring." In I. Neumann-Holzschuh and E.W. Schneider (eds), 41-63.

REFERENCES

311

Baker, P. (ed.). 1995. From Contact to Creole and Beyond [Westminster Creolistics Series 1]. London: University of Westminster Press. Baker, P. and Bruyn, A. (eds). 1999. St Kitts and the Atlantic creoles [Westminster Creolistics Series 4]. London: University of Westminster Press. Baker, P. and Corne, C. 1982. Isle de France Creole. Ann Arbor, MI: Karoma. Baker, P. and Syea, A. 1991. "On the copula in Mauritian Creole, past and present." In F. Byrne and T. Huebner (eds), 159-175. Baker, P. and Syea, A. (eds). 1996. Changing Meanings, Changing Functions [Westminster Creolistics Series 2]. London: University of Westminster Press. Bakker, P. 1987. "Reduplications in Saramaccan." In Studies in Saramaccan Language Structure [Caribbean Culture Studies 2], M.C. Alleyne (ed.), 17-40. Amsterdam: Universiteit van Amsterdam. Bakker, P. 1989. "Relexification in Canada: The case of Métif (French-Cree)." In C. Lefebvre and J.S. Lumsden (eds), 339-350. Bakker, P. 1992. "A language of our own: The genesis of Mitchif, the mixed Cree-French language of the Canadian Metis." Unpublished Ph.D. dissertation, University of Amsterdam, Amsterdam. Bakker, P. 1994. "Michif, the Cree-French mixed language of the Métis Buffalo Hunters in Canada." In P. Bakker and M. Mous (eds), 13-33. Bakker, P. and Mous, M. (eds). 1994. Mixed Languages [Studies in Language and Language Use 13]. Dordrecht: ICG Printing. Bakker, P., Smith, N. and Veenstra, T. 1995. "Saramaccan." In J. Arends, P.C. Muysken, and N. Smith (eds), 165-178. Bámgbósé, A. 1974. "On serial verbs and verbal status." The Journal of West African Languages 9(1): 17-48. Bastide, R. 1967. Les Amériques noires. Paris: Payot. Bauer, L. 2001. Morphological Productivity. Cambridge: Cambridge University Press. Baugh, J. 1999. Out of the Mouths of Slaves: African American Language and Educational Malpractice. Austin, TX: University of Texas Press. Bennis, H. 1982. "De Pro-drop-Parameter en Subjektloze Zinnen in het Nederlands." Spektador 6: 4 0 9 - 2 7 . Bentolila, A. 1971. Les systèmes verbaux créoles : comparaison avec les langues africaines. Unpublished Ph.D. dissertation, Université Paris V (René Descartes), Paris. Bentolila, A. 1987. "Marques aspecto-temporelles en créole haïtien : de l'analyse synchronique à la formulation d'hypothèses diachroniques." La Linguistique 23 (1): 103-122. Bickerton, D. 1977. "Pidginisation and creolisation: Language acquisition and language universals." In A. Valdman (ed.), 49-69. Bickerton, D. 1981. Roots of Language. Ann Arbor, MI: Karoma. Bickerton, D. 1984. "The language bioprogram hypothesis." The Behavioral and'Brain Sciences 7 (2): 173-221. Bickerton, D. 1986. "Creoles and West African languages: A case of mistaken identity?" In P.C. Muysken and N. Smith (eds), 25-40.

312

ISSUES IN THE STUDY OF PIDGIN AND CREOLE LANGUAGES

Bickerton, D. 1988. "Creole languages and the bioprogram." In F.J. Newmeyer (ed.), 268-284. Bickerton, D. 1996. "Why serial verb construction in Isle de France Creole can have subjects: A reply to Corne, Coleman and Curnow." In P. Baker and A. Syea (eds), 155-169. Bloomfield, L. 1933. Language. New York, NY: Holt, Rinehart, and Winston. Bloomfield, L. (1887-1949). A Leonard Bloomfield Anthology. C.F. Hockett (ed.) (1970) (Indiana University Studies in the History and Theory of Linguistics). Bloomington, IN: Indiana University Press. Bolé-Richard, R. 1983. Systématique phonologique et grammaticale d'un parlé ewe : le gen-mina du Sud-Togo et Sud-Bénin. Paris: L'Harmattan. Booij, G. and van Marie, J. (eds). 2003. Yearbook of Morphology 2002. Dordrecht: Kluwer Academic Publishers. Boretzky, N., Enninger, W. and Stolz, T. (eds). 1988. Bochum-Essener Beiträge zur Sprachwandelforschung. Bochum: Studienverlag Brockmeyer. Botha, R.P. 1976. ""Gut feelings" in generative grammar." Bloomington, IN: Indiana University Linguistics Club [Brochure]. Bouchard, D. 1982. "Les constructions relatives en français vernaculaire et en français standard : étude d'un paramètre." In C. Lefebvre (ed.), 103-133. Bouchard, D. 1995. The Semantics of Syntax. Chicago, IL: University of Chicago Press. Bouquiaux, L. and Thomas, J.M.C. 1992. Studying and Describing Unwritten Languages. Dallas, TX: Summer Institute of Linguistics. Braun, M. and Plag, I. 2003. "How transparent is creole morphology? A study of Early Sranan word-formation." In G. Booij and J. van Marie (eds). Bresnan, J. 1979. Theory of Complementation in English Syntax. New York, NY: Garland. Brousseau, A.-M. 1988. "Triptyque sur les composés : les noms composés en français, fongbe et haïtien en regard des notions de tête et de percolation." Travaux de recherche sur le créole haïtien 2: 1-109 [Université du Québec à Montréal]. Brousseau, A.-M. 1989. "De 'nù-fló' à 'po-bouche' : hypothèses sur l'origine des composés en haïtien." In C. Lefebvre and J.S. Lumsden (eds), 285-312. Brousseau, A.-M. 1990. "Panorama de la morphologie du Fongbe." Journal of West African Languages 20 (1): 27-45. Brousseau, A.-M. 1993. "Représentations sémantiques et projections syntaxiques des verbes en fongbè." Unpublished Ph.D. dissertation, Université du Québec à Montréal, Montreal. Brousseau, A.-M. 1994. "Morphological structure and relexification." In C. Lefebvre and J.S. Lumsden (eds) (1994b), 30 pages. Brousseau, A.-M. 1995a. "Les pronoms en créole haïtien, en français et en fongbé." In C. Lefebvre (ed.), vol. 4, 33 pages. Brousseau, A.-M. 1995b. "Réflexivité et classes de verbes en créole haïtien et en fongbé." In C. Lefebvre (ed.), vol. 4, 15 pages. Brousseau, A.-M. 1998. Réalisations argumentales et classes de verbes en fongbè [Langues et cultures africaines 22]. Paris: Peeters.

REFERENCES

313

Brousseau, A.-M. 1998. Réalisations argumentales et classes de verbes en fongbè [Langues et cultures africaines 22]. Paris: Peeters. Brousseau, A.-M. In preparation. The Genesis of the Phonological System of Haitian Creole [provisional title]. University of Toronto. Brousseau, A.-M., Filipovich, S. and Lefebvre, C. 1989. "Morphological processes in Haitian Creole: The question of substratum and simplification." Journal of Pidgin and Creole Languages 4(1): 1-36. Brousseau, A.-M. and Lumsden, J.S. 1992. "Nominal structures in Fongbe." Topics on the Syntax and Semantics of Fongbe. Special issue of The Journal of West African Languages 22 (1): 5-26. Brousseau, A.-M. and Nikiema, E. 2001. Phonologie et morphologie du français [Collection champs linguistiques]. Montreal: Fides. Brunot, F. 1905. Histoire de la langue française des origines à 1900. Paris: A. Colin. Brunot, F. 1926. La pensée et la langue. Paris: Masson et cie. Bruyn, A. 1995. Grammaticalization in Creoles: The Development of Determiners and Relative Clauses in Sranan [Studies in Language and Language use 21]. Amsterdam: The Netherlands. Bruyn, A. 1996. "On identifying instances of grammaticalisation in creole languages." In P. Baker and A. Syea (eds), 29-47. Burling, R. 1984. Learning a Field Language. Ann Arbor, MI: University of Michigan Press. Byrne, F. 1987. Grammatical Relations in a Radical Creole. Verb Complementation in Saramaccan [Creole Language Library 3]. Amsterdam/Philadelphia: John Benjamins Publishing Company. Byrne, F. and Holm, J. (eds). 1993. Atlantic Meets Pacific: A Global View of Pidginization and Creolization. Selected Papers from the Society for Pidgin and Creole Linguistics [Creole Language Library 11]. Amsterdam/ Philadelphia: John Benjamins Publishing Company. Byrne, F. and Huebner, T. (eds). 1991. Development and Structures of Creole Languages. Essays in Honor of Derek Bickerton [Creole Language Library 9]. Amsterdam/Philadelphia: John Benjamins Publishing Company. Byrne, F. and Winford, D. (eds). 1993. Focus and Grammatical Relations in Creole Languages [Creole Language Library 12]. Amsterdam/Philadelphia: John Benjamins Publishing Company. Cadely, J.-R. 1994. "Aspects de la phonologie du créole haïtien." Unpublished Ph.D. dissertation, Université du Quebec à Montréal, Montreal. Campbell, L. 1985. The Pipil Language of El Salvador [Mouton Grammar Library 1]. Berlin/New York: Mouton de Gruyter. Capo, H.B.C. 1984. "Elements of Ewe-Gen-Aja-Fon dialectology." In Peuples du golfe du Bénin : Aja-Ewé [Colloque de Cotonou], F. de Medeiros (ed.), 167-178. Paris: Karthala. Capo, H.B.C. 1991. A comparative phonology of gbe. Dordrecht: Foris Publications. Carden, G. and Stewart, W.A. 1988. "Binding theory, bioprogram, and creolisation: Evidence from Haitian Creole." Journal of Pidgin and Creole Languages 3 (1): 1-68.

314

ISSUES IN THE STUDY OF PIDGIN AND CREOLE LANGUAGES

Cattell, R. 1984. Composite Predicates in English [Syntax and Semantics 17]. North Ryde, Australia: Academic Press. Chansou, M. 1997. "The methodology of corpus building." Terminologies nouvelles 16 (2): 33-36. Chaudenson, R. 1973. "Pour une étude comparée des créoles et parlers français d'outre-mer : survivance et innovation." Revue de linguistique romane 147/148: 342-371. Chaudenson, R. 1977. "Toward the reconstruction of the social matrix of creole language." In A. Valdman (ed.), 259-277. Chaudenson, R. 1979. Les créoles français. Paris: Nathan. Chaudenson, R. 1983. "Où l'on reparle de la genèse et des structures des créoles de l'Océan Indien." Études créoles 6 (2): 157-224. Chaudenson, R. 1992. Des îles, des hommes, des langues : Essai sur la créolisation linguistique et culturelle. Paris: L'Harmattan. Chaudenson, R. 1993. "De l'hypothèse aux exemples. Un cas de créolisation : selon la formation des systèmes de démonstratifs créoles." Études créoles 16 (1): 17-38. Chaudenson, R. 1994. "Comparatisme et méthode générative transformationnelle : le cas des études sur le créole haïtien." Créolistique et grammaire générative, Pluralismes 8: 1-20. Chaudenson, R. 1996a. "Petit sottisier glané dans les recherches sur l'origine des systèmes grammaticaux créoles." In Matériaux pour I'étude des classes grammaticales dans les langues créoles, D. Véronique (ed.), 17-31. Aix-enProvence: Publication de l'Université de Provence. Chaudenson, R. 1996b. "Démystification de la relexification." Études créoles 19 (1): 93-109. Chaudenson, R. 2000. "Créolisation du français et francisation du créole : les cas de Saint-Barthélemy et de la Réunion." In I. Neumann-Holzschuh and E.W. Schneider (eds), 361-381. Chomsky, N. 1972. Studies on Semantics in Generative Grammar. The Hague: Mouton de Gruyter. Chomsky, N. 1974. "The Amherst Lectures." Unpublished Ms. Chomsky, N. 1981. Lectures on Government and Binding. Dordrecht: Foris Publications. Chomsky, N. 1986. Knowledge of Language: Its Nature, Origin and Use. New York, NY: Praeger. Clark, E. and Clark, H. 1979. "When nouns surface as verbs." Language 55 (4): 767-811. Clements, G.N. 1975. "The logophoric pronoun in Ewe: Its role in discourse." Journal of West African Languages 10 (2): 141-177. Clements, J.C. 1996. The Genesis of a Language: The Formation and Development of Korlai Portuguese [Creole Language Library 16]. Amsterdam/Philadelphia: John Benjamins Publishing Company. Collins, C. 1994. "The factive construction in Kwa." Travaux de recherche sur le créole haïtien 23: 31-65 [Université du Québec à Montréal]. Corbin, D. 1987. Morphologie dérivationnelle et structuration du lexique [2 volumes]. Lille: Presses Universitaires de Lille.

REFERENCES

315

Corne, C. 1977. Seychelles Creole Grammar. Tübingen: TBL Verlag Gunter Narr. Corne, C. 1979. "Le français à Tahiti." In A. Valdman (ed.), 631-661. Corne, C. 1986. "Universals, substrata, and the Indian Ocean creoles." In P.C. Muysken and N. Smith (eds), 163-168. Corne, C. 1995. "A contact-induced and vernacularized language: How Melanesian is Tayo?" In P. Baker (ed.), 121-148. Corne, C. 1999. From French to Creole. The Development of New Vernaculars in the French Colonial World [Westminster Creolistics Series 5]. London: University of Westminster Press. Cowart, W. 1998. Experimental Syntax. Applying Objective Methods to Sentence Judgments. Beverly Hills/London: Sage. Cowper, E.A. 1989. 'Thematic underspecification: The case of have." The Canadian Linguistics Association Annual Meeting, Université de Laval, Quebec City. Cowper, E.A. 1995. "English participle construction." The Canadian Journal of Linguistics 40 (1): 1-38. Crowley, T. In press. "Pidgin/creole morphology." In The Handbook of Pidgins and Creoles, J. Singler and S. Kouwenberg (eds). Oxford: Blackwell. Curtin, P.D. 1976. "The Atlantic Slave Trade, 1600-1800." In History of West Africa, J.F. Ade Ajayi and M. Crowder (eds), 302-330. London: Longman. Da Cruz, M. 1994 "Contribution à l'étude de la négation en Fongbè." In C. Lefebvre and J.S. Lumsden (eds) (1994b), Volume III, 69-111. Damoiseau, R. 1988. "Éléments pour une classification des verbaux en créole haïtien." Études créoles 11 (1): 41-64. Damoiseau, R. 1991. "Exemples de procédures de réinterprétation d'un lexique français en créole haïtien." Études créoles 14 (1): 30-43. DeCamp, D. 1974. "Neutralisations, iteratives, and ideophones: The locus of language in Jamaica." In Pidgins and Creoles, D. DeCamp and I. Hancock (eds), 42-60. Washington, DC: Georgetown University Press. Déchaine, R.-M. 1988. "Opérations sur les structures d'arguments : Le cas des constructions sérielles en Haïtien." Unpublished MA thesis, Université du Québec à Montréal, Montreal. Déchaine, R.-M. 1991. "Bare sentences." In Proceedings of SALT I, S. Moore and A. Wyner (eds), 31-50. Cornell, NY: Cornell University Working Papers in Linguistics. Déchaine, R.-M. 1993. "Predicates across categories." Unpublished Ph.D. dissertation, University of Massachusetts, Amherst, MA. Déchaine, R.-M. and Manfredi, V. 1994. "Binding domains in Haitian." Natural Language and Linguistic Theory 12 (2): 203-257. Dees, A. 1971. Étude sur l'évolution des démonstratifs en ancien et moyen français. Groningen: Wolters-Noordhoff. DeGraff, M.A.F. 1992a. "Is Haitian Creole a pro-drop language?" Travaux de recherche sur le créole haïtien 11: 1-22 [Université du Québec à Montréal].

316

ISSUES IN THE STUDY OF PIDGIN AND CREOLE LANGUAGES

DeGraff, M.A.F. 1992b. "Creole grammars and the acquisition of syntax: The case of Haitian." Unpublished Ph.D. dissertation, University of Pennsylvania, Philadelphia, PA. DeGraff, M.A.F. 1992c. "The syntax of predication in Haitian." In Proceedings of the 22nd Annual Meeting of the Northeastern Linguistics Society, Amherst, MA: University of Massachusetts, Graduate Linguistics Students Association. DeGraff, M.A.F. 1992d. "Haitian null subjects revisited." Travaux de recherche sur le créole haïtien 11: 59-74 [Université du Québec à Montréal]. DeGraff, M.A.F. 1993a. "A riddle on negation in Haitian." Probus 5 (1): 63-93. DeGraff, M.A.F. 1993b. "Is Haitian Creole a pro-drop language." In F. Byrne and J. Holm (eds), 71-90. DeGraff, M.A.F. 1994. "The morphology-syntax interface in creolisation (and diachrony)." Studies in the Linguistic Sciences 24 (1): 115-131. DeGraff, M.A.F. 1996. "Creole languages and parameter setting: A case study using Haitian Creole and the pro-drop parameter." In H. Wekker (ed.), 65-105. DeGraff, M.A.F. 1999. "Empirical quicksand: Probing two recent articles on Haitian Creole." Journal of Pidgin and Creole Languages 14 (2): 359-377. DeGraff, M.A.F. 2001. "Morphology in creole genesis: A prolegomenon." In Ken Hale: A life in language, M. Kenstowicz (ed.), 53-121. Cambridge, MA: The MIT Press. DeGraff, M.A.F. (ed.). 1999. Language Creation and Language Change. Cambridge, MA: The MIT Press. De Guiraudon, T.G. 1894. Manuel de la langue foule. Paris/Leipzig: H. Welter. Dejean, Y. 1993. Comments from M. Yves Dejean on the articles published in Travaux de recherche sur le créole haïtien (Addressed to me by the author), Université du Québec à Montréal. Delafosse, M. 1929. La langue mandingue et ses dialectes (malinke, bombara, dioula). Paris: Geuthner. Denis, M.-J. In preparation, "-syon, -ès et -te en créole haïtien : des affixes productifs ou non productifs?" Unpublished MA thesis, Université du Québec à Montréal, Montreal. Déprez, V. 1992a. "Is Haitian Creole really a pro-drop language?" Travaux de recherche sur le créole haïtien 11: 23-40 [Université du Québec à Montréal]. Déprez, V. 1992b. "Raising constructions in Haitian Creole." Natural Language and Linguistic Theory 10 (2): 191-231. Déprez, V. 1994. "Haitian Creole: A pro-drop language?" Journal of Pidgin and Creole Languages 9 (1): 1-24. Déprez, V. 1997. "Two types of negative concord." Probus 9 (2): 103-142. Déprez, V. 1999. "The roots of negative concord in French and French-lexicon creoles." In M.A.F. DeGraff (ed.), 375-427. Déprez, V. and Vinet, M.-T. 1991. " "Se" in Haitian Creole: an ambiguous X' category." Unpublished Ms.

REFERENCES

317

Diessel, H. 1999. Demonstratives: Form, Function and Grammaticalization [Typological Studies in Language 42]. Amsterdam/Philadelphia: John Benjamins Publishing Company.. Dijkhoff, M. 1993. "Papiamentu word formation: A case study of complex nouns and their relation to phrases and clauses." Unpublished Ph.D. dissertation, University of Amsterdam, Amsterdam. Dillard, J.L. 1970. "Principles in the history of American English: Paradox, virginity, and cafeteria." Florida FL Reporter 7 (Spring/Fall): 32-33. Di Sciullo, A.-M. and Williams, E. 1987. On the Definition of Word [Linguistic Inquiry Monograph Series 14]. Cambridge, MA: The MIT Press. Domingue, N. 1980. "Syntactic innovations in Mauritian Bhojpuri." Unpublished Ms., McGill University, Montreal. Domingue, N. 1981. "Internal change in a transplanted language." Studies in the Linguistic Sciences 4 (2): 151-159. Donaldson, B.C. 1993. A Grammar of Afrikaans [Mouton Grammar Library 8]. Berlin/New York: Mouton de Gruyter. Donohue, M. 1999a. "Syntactic categories in Tukang Besi." Revue québécoise de linguistique 27 (2): 71-90. Donohue, M. 1999b. A Grammar of Tukang Besi [Mouton Grammar Library 20]. Berlin/New York: Mouton de Gruyter. Doran, M., Drapeau, L. and Lefebvre, C. 1982. "Le projet CentreSud : Corpus et méthodologie." In C. Lefebvre (ed.), 73-125. Dubois, J. 1962. Étude sur la dérivation suffixale en français moderne et contemporain. Paris: Librairie Larousse. Dumais, D. 1988. "INFL en créole haïtien." In Études syntaxiques, morphologiques et phonologiques, C. Lefebvre (ed.), 242-265. Research report prepared for SSHRCC, FCAR and PAFAC on the project Haiti-Fon. Université du Québec à Montréal, Montreal. Duponchel, L. and Mel, G. 1983. "L'alladian." In G. Hérault (ed.), 181-225. Durie, M. In preparation. Language, Function and Time. London: Oxford University Press. Escure, G. 1993. "Focus, topic particles and discourse markers in the Belisean creole continuum." In F. Byrne and D. Winford (eds), 233-247. Escure, G. 1997. Creole and Dialect Continua. Standard Acquisition Processes in Belize and China (PRC) [Creole Language Library 18]. Amsterdam/Philadelphia: John Benjamins Publishing Company. Étienne, G. 1974. "Le créole du nord d'Haïti : Étude des niveaux de structure." Unpublished Ph.D. dissertation, Université des sciences humaines de Strasbourg, Strasbourg. Faine, J. 1937. Philologie créole : études historiques et étymologiques sur la langue créole d'Haïti. Port-au-Prince: Imprimerie de l'Etat. Faine, J. 1974. Dictionnaire français-créole. Ottawa: Leméac. Faltz, L.M. 1985. Reflexivization: A Study in Universal Syntax. New York/London: Garland. Fattier, D. 1998. "Contribution à l'étude de la genèse d'un créole : l'Atlas linguistique d'Haïti, cartes et commentaires." Unpublished Ph.D. dissertation, Université de Provence, Marseilles.

318

ISSUES IN THE STUDY OF PIDGIN AND CREOLE LANGUAGES

Fauchois, A. 1983. "Opposition verbo-nominales : Reflexions sur le créole haïtien." Unpublished Ms., Université Paris V (René Descartes). Férère, G.A. 1974. "Haitian Creole sound-system, form-classes, texts." Unpublished Ph.D. dissertation, University of Pennsylvania, Philadelphia, PA. Ferguson, C.A. 1971. "Absence of copula and the notion of simplicity: A study of normal speech, baby talk, foreigner talk, and pidgins." In D. Hymes (ed.), 141-150. Ferraz, L. 1983. "The origin and development of four creoles in the Gulf of Guinea in the social context of creolization." In E. Woolford and W. Washabaugh (eds), 120-125. Filipovich, S. 1987. "La morphologie de l'haïtien." Unpublished MA thesis, Université du Québec à Montréal, Montreal. Fillmore, C.J. 1975. Santa Cruz Lectures on Deixis 1971. Bloomington, IN: Indiana University Linguistics Club. Foley, W.A. 1988. "Language birth: The processes of pidginization and creolization." In F.J. Newmeyer (ed.), 162-183. Fournier, R. 1977. "N ap fè yun ti-kose su la (La grammaire de la particule la en créole haïtien)." Unpublished MA thesis, Université du Québec à Montréal. Frake, C.O. 1971. "Lexical origins and semantic structure in Philippine Creole Spanish." In D. Hymes (ed.), 223-243. Freeman, B.C. 1988. Dictionnaire inverse de la langue haïtienne. Lawrence, KS: Institute of Haitian Studies, The University of Kansas. Freeman, B.C. and Laguerre, J. 1996. [1998] Haitian-English Dictionary (2nd ed. 1998). Lawrence, KS: Institute of Haitian Studies, The University of Kansas. Gacic, M. 1994. "The methodology of compiling a terminological dictionary." Filologija 22-23: 375-386. Gambhir, S.K. 1988. "Structural development in Guyanese Bhojpuri." In Language Transplanted: The Development of Overseas Hindi, R. Barz and J. Siegel (eds), 69-94. Wiesbaden: Harrassowitz. Germain, R. 1980. Grammaire créole. Paris: l'Harmattan. Ghomeshi, J. and Massam, D. 1994. "Lexical/syntactic relations without projection." Linguistic Analysis 24 (3-4): 175-217. Gilbert, G.G. 2002. "What's ahead in pidgin and creole linguistics." In G.G. Gilbert (ed.), 1-6. Gilbert, G.G. (ed.). 2002. Pidgin and Creole Linguistics in the 21st Century [Studies in Ethnolinguistics 9]. New York, NY: Peter Lang Publishing. Gilles, R. 1988. "Réalisation du cas en créole haïtien dans trois environnements : les prépositions, les marqueurs de cas et les noms." Unpublished MA thesis, Université du Québec à Montréal, Montreal. Gilman, C. 1993. "A Bantu model for the Seychellois pour-dire complementizer." In F. Byrne and J. Holm (eds), 49-55. Givón, T. 1971. "On the verbal origin of the Bantu suffixes." Studies in African Linguistics 2: 145-163. Göbl-Galdi, L. 1934. "La structure grammaticale des patois français-créoles." Zeitschrift für französische Sprache und Literatur 58: 257-295.

REFERENCES

319

Goddard, Y. 1987. "Leonard Bloomfield's descriptive and comparative studies of Algonquian." Historiographia Linguistica 14 (1-2): 179-217. Goldsmith, J. 1978. "'Que' c'est quoi? 'Que', c'est 'quoi'." Recherches linguistiques à Montréal 11: 1-15. Goodman, M.F. 1964. A Comparative Study of Creole French Dialects. The Hague: Mouton de Gruyter. Goodman, M.F. 1971. 'The strange case of Mbugu (Tanzania)." In D. Hymes (ed.), 243-255. Goodman, M.F. 1985. "Review of Roots of Language, by Derek Bickerton." International Journal ofAmerican Linguistics 51 (1): 109-137. Goodman, M.F. 1999. "On the origin of Hawaiian Creole English: A rejoinder to Roberts." Language 75 (2): 347-348. Gougenheim, G. 1973. Grammaire de la langue française au seizième siècle. Paris: A. et J. Picard. Grenoble, L. and Riley, M. 1996. "The role of deictics in discourse coherence: French voici/voilà and Russian votlvon" Journal of Pragmatics 25 (6): 819-838. Grevisse, M. 1975. Le bon usage [10th edition]. Gembloux, Belgium: Duculot. Grimes, B.F. (ed.) 2000. Ethnologue. Languages of the word [14th edition]. Dallas, TX: Summer Institute of Linguistics. Grimshaw, J. 1971. "Some social forces and some social functions of pidgin and creole languages." In D. Hymes (ed.), 427-447. Grimshaw, J. and Meister, A. 1988. "Light verbs and 9-marking." Linguistic Inquiry 19 (2): 205-232. Guiraud, P. 1966. Le Moyen Français [2nd edition] [Que sais-je?]. Paris: Presses universitaires de France. Gundel, J.K. Hedberg, N. and Zacharski, R. 1993. "Cognitive status and the form of referring expressions in discourse." Language 69: 274-307. Haase, A. 1975. Syntaxe française du XVIIe siècle [5th edition]. Paris: Delagrave. Hagège, C. 1974. "Les pronoms logophoriques." Bulletin de la Société de linguistique de Paris 1: 287-310. Hagège, C. 1985. L'homme de parole : contribution linguistique aux sciences humaines. Paris: Librairie Arthème Fayard. Hale, K. and Keyser, J. 1987. A view from the middle [MIT Lexicon Project Working Papers 10]. Cambridge, MA: MIT Center for Cognitive Science. Hale, K. and Keyser, J. 1993. "On argument structure and the lexical expression of syntactic relations." In The view from building 20, K. Hale and J. Keiser (eds), 53-111. Cambridge, MA: The MIT Press. Hall, R.A. 1950. "The genetic relationships of Haitian Creole." Ricerche Linguistiche 1: 194-203. Hall, R.A. 1953. Haitian Creole: Grammar, Texts and Vocabulary [American Anthropologist Association 74]. Menasha, WI: Georges Banta Publishing Company. Hall, R.A. 1958. "Creolized languages and 'genetic relationships'." Word 14: 367-373.

320

ISSUES IN THE STUDY OF PIDGIN AND CREOLE LANGUAGES

Hall, R.A. 1966. Pidgin and Creole Languages. Ithaca, NY: Cornell University Press. Hancock, I.F. 1968. "A provisional comparaison of the English-based Atlantic creoles, with accompanying word lists." Paper presented at the Conference of Pidgin and Creole Languages, Mona, Jamaica, April. Hancock, I F . 1971. "A map and list of pidgin and creole languages." In D. Hymes (ed.), 509-523. Hancock, I.F. 1980. "Gullah and Barbadian: Origins and relationships." American Speech 55: 17-35. Hancock, I.F. 1987. "A preliminary classification of the anglophone Atlantic creoles with syntactic data from thirty-three representative dialects." In Pidgin and Creole Languages: Essays in Memory of John E. Reinecke, G.G. Gilbert (ed.), 264-333. Honolulu, HI: University of Hawaii Press. Harris, J.W. 1991. "Kriol-the creation of a new language." In Language in Australia, S. Romaine (ed.), 195-203. Cambridge: Cambridge University Press. Haspelmath, M. 1998. "Does grammaticalization need reanalysis?" Studies in Language 22 (2): 315-351. Haspelmath, M. 1999. "Why is grammaticalization irreversible?" Linguistics 37 (6): 1043-1068. Hay, J. 2001. "Lexical frenquency in morphology: Is everything relative?" Linguistics 39 (6): 1041-1070. Hazaël-Massieux, G. 1996. Les créoles : Problèmes de genèse et de description. Aix-en Provence: Publications de l'Université de Provence. Hazoumê, M.L. 1990. Essai de classification synchronique. Étude comparative des parlers Gbe du Sud-Bénin. Cotonou, Benin: Centre national de linguistique appliquée. Heine, B., Claudi, U. and Hünnemeyer, F. 1991. Grammaticalization. A Conceptual Framework. Chicago, IL: University of Chicago Press. Heine, B. and Reh, M. 1984. Grammaticalisation and Reanalysis in African Languages. Hamburg: Helmut Buske Verlag. Hérault, G. 1983a. "L'adioukrou." In G. Hérault (ed.), 129-153. Hérault, G. 1983b. "L'eotile." In G. Hérault (ed.), 403-424. Hérault, G. (ed.). 1983. Atlas des langues kwa de Côte d'Ivoire [vol. 1, 2nd edition]. Paris: Agence de coopération culturelle et technique/Abidjan, Ivory Coast: Université d'Abidjan, Institut de linguistique appliquée. Herskovits, M.J. 1975. Life in a Haitian Valley. New York, NY: Octagon Books. Hesseling, D.C. 1933. "Een Spaans boek over het Papiaments." TNTL 52: 40-57; review of Rodolfo Lenz El Papiamento, la lengua criolla de Curazao (la grammatica mas sencilla), Anales de la Universidad de Chile, Seperatum (Santiago, 1928). Hilaire, J. 1992. Soubassements amérindiens [L'édifice créole en Haïti 1]. Fribourg, Switzerland: Edikreyol. Hilaire, J. 1993. Soubassements africains [L'édifice créole en Haïti 2]. Fribourg, Switzerland: Edikreyol.

REFERENCES

321

Hinskens, F. and van Rossem, C. 1995. "The negerhollands word sender in eighteenth-century manuscripts." In J. Arends (ed.), 63-87. Hirschbühler, P. 1978 "The Semantics and Syntax of Wh-Constructions." Unpublished Ph.D. dissertation, University of Massachusetts, Amherst, MA. Holm, J. 1988. Pidgins and Creoles [Cambridge Language Surveys 1]. Cambridge: Cambridge University Press. Holm, J. 2000. "Semi-creolization: Problems in the development of theory." In I. Neumann-Holzschuh and E.W. Schneider (eds), 19-40. Hopper, P.J. and Traugott, E.C. 1993. Grammaticalization [Cambridge Textbooks in Linguistics]. Cambridge: Cambridge University Press. Huber, M. and Parkvall, M. (eds). 1999. Spreading the Word. The Issue of Diffusion Among the Atlantic Creoles [Westminster Creolistics Series 6]. London: University of Westminster Press. Hulk, A. 1986. "Subject clitics and the PRO-drop parameter." In Formal parameters of generative grammar, P. Coopmans, I. Bordelois and B.D. Smith (eds), 107-120. Dordrecht: ICG Printing. Hull, A. 1979. "Affinités entre les variétés du français." In A. Valdman (ed.), 165-182. Huttar, G.L. 1975. "Sources of creole semantic structures." Language 51 (3): 684-695. Huttar, G.L. and Koanting, E. 1993. "Are Ndjuká comparative markers verbs?" In F. Byrne and J. Holm (eds), 165-174. Hyams, N.M. 1986. Language Acquisition and the Theory of Parameters. Dordrecht: Reidel. Hyams, N.M. 1987. "Theory of parameters and syntactic development." In Parameter Setting, T. Roeper and E. Williams (eds), 1-22. Dordrecht: Reidel. Hyman, L.M. and Comrie, B. 1981. "Logophoric reference in Gokana." Journal ofAfrican Languages and Linguistics 3 (1): 19-37. Hymes, D. 1971. "Introduction." In D. Hymes (ed.), 65-91. Hymes, D. (ed.). 1971. Pidginisation and Creolisation of Languages [Proceedings of a conference held at the University of West Indies, Mona, Jamaica, April 1968]. Cambridge: Cambridge University Press. Isquerdo, A.N. 1998. "The issue of the constitution of the corpus: Limits and possibilities." Estudos Linguísticos 27: 272-277. Issoufi, O.A. 1998. "L'enquête sur les pratiques et les représentations linguistiques." Cahiers de l'Institut de linguistique et des sciences du language (Université de Lausanne) 10: 141-154. Jackendoff, R. 1976. "Toward an explanatory semantic representation." Linguistic Inquiry 7 (1): 89-150. Jackendoff, R. 1977. X' Syntax: A Study of Phrase Structure [Linguistic Inquiry Monograph 2]. Cambridge, MA: The MIT Press. Jaeggli, O.A. 1984. "Subject extraction and the null subject parameter." In C. Jones and P. Sells (eds), 132-153. Jaggar, P.J. and Buba, M. 1994. 'The space and time adverbials nan/can in Hausa: Cracking the deictic code." Language Sciences 16 (3-4): 387-421.

322

ISSUES IN THE STUDY OF PIDGIN AND CREOLE LANGUAGES

Jensen, B., Koopman, H. and Muysken, P.C. 1978. "Serial verbs in the creole languages." Amsterdam Creole Studies 11 (2): 125-159. Jespersen, O. 1922. Language: Its Nature, Development and Origin. London: Allen and Unwin. Johns, A. 1992. "Deriving ergativity." Linguistic Inquiry 23 (1): 57-87. Johnson, K. 1991. "Object positions." Natural Language and Linguistic Theory 9: 577-639. Jones, C. and Sells, P. (eds). 1984. Proceedings of NELS 14. Amherst, MA: University of Massachusetts at Amherst, GLSA. Joseph, F. 1988. "La détermination nominale en créole haïtien." Unpublished Ph.D. dissertation, Université Paris VII (Denis Diderot), Paris. Joseph, J.S. 1994. "De l'étude sémantico-syntaxique de la locution nan du créole haïtien." Unpublished MA thesis, Université du Québec à Montréal, Montreal. Joseph, J.S. 1995. "La négation en créole Haïtien: un cas de relexification." Université du Québec à Montréal, Montreal, ms. Joseph, J.S. 1999. "Étude semantico-syntaxique des adjectifs et des adverbes en créole haïtien." Unpublished Ph.D. dissertation, Université du Québec à Montréal, Montreal. Jourdan, C. 1985. "Sapos Yumi Mitim Yumi: Urbanization and creolization of Solomon Islands Pidgin." Unpublished Ph.D. dissertation, Australian National University, Canberra. Jourdan, C. and Keesing, R.M. 1997. "From Pisin to Pidgin: Creolization in process in the Salomon Islands." Language and Society 26 (3): 401-420. Jourdan, C. and Tuite, K. (eds). 2003. Ethnolinguistics and Anthropological Theory [Collection champs linguistiques]. Montreal: Fides. Julliand, A. 1965. Dictionnaire inverse de la langue française. The Hague: Mouton de Gruyter. Juneau, M. 1972. Contribution à l'histoire de la prononciation française au Québec. Pari s/Sainte-Foy: Librairie C. Klincksieck/Presses de l'Université Laval. Juneau, M. and Poirier, C. 1973. Le Livre de comptes d'un meunier québécois (fin XVIIe-début XVIIIe siècle). Sainte-Foy: Presses de l'Université Laval. Kay, P. and Sankoff, G. 1974. "A language-universals approach to pidgins and creoles." In Pidgins and Creoles. Current Trends and Prospect, D. DeCamp and I. Hancock (eds), 61-72. Washington, DC: Georgetown University Press. Kayne, R.S. 1976. "French relative que." In Current Studies in Romance Linguistics, F.G. Hensey and M. Lujan (eds), 255-299. Washington, DC: Georgetown University Press. Kayne, R.S. 1978. "Stylistic inversion, successive cyclicity, Move NP in French." Linguistic Inquiry 9 (4): 592-621. Kayne, R.S. 1984. Connectedness and Binary Branching. Dordrecht: Foris Publications. Keesing, R.M. 1988. Melanesian Pidgin and the Oceanic Substrate. Stanford, CA: Stanford University Press.

REFERENCES

323

Keesing, R.M. 1991. "Substrates calquing and grammaticalization in Melanesian pidgin." In E. Traugott and B. Heine (eds), 315-342. Kibrik, A.K. 1977. The Methodology of Field Investigation in Linguistics. The Hague: Mouton de Gruyter. Kihm, A. (ed.). 1991. La Créolisation : Théorie et applications [Recherches linguistiques de Vincennes 20]. Paris: Presses universitaires de Vincennes. Kihm, A. 1994. Kriyol Syntax. The Portuguese-Based Creole Language of Guinea-Bissau [Creole Language Library 14]. Amsterdam/Philadelphia: John Benjamins Publishing Company. Kinyalolo, K.K.W. 1992. "A note on word order in the progressive and prospective in Fon." Topics on the Syntax and Semantics of Fongbe. Special issue of The Journal of West African Languages 22 (1): 35-51. Kinyalolo, K.K.W. 1993a. "Conditionals and factives in Fon." In Études syntaxiques, vol. 4, C. Lefebvre and J.S. Lumsden (eds), 161-173. Research report prepared for SSHRCC, FCAR and PAFAC on the project La genèse du créole haïtien : un cas particulier d'investigation sur la forme de la grammaire universelle [4 volumes]. Université du Québec à Montréal, Montreal. Kinyalolo, K.K.W. 1993b. "On some syntactic properties of ɖò in Fon." Lingua 91 (2-3): 201-233. Kinyalolo, K.K.W. 1993c. "The logophoric pronoun émi as a LF operatoranaphor." In Proceedings of NELS 23, A. Schafer (ed.), 223-237. Amherst, MA: GLSA. Kinyalolo, K.K.W. 1994. "Pronouns and relexification." In C. Lefebvre and J.S. Lumsden (eds) (1994b), vol. 3, 259-293. Kiple, K. 1984. The Caribbean Slave: A Biological History. Cambridge: Cambridge University Press. Koopman, H. 1982a. "Les constructions relatives." In C. Lefebvre, H. Magloire-Holly and N. Piou (eds), 167-203. Koopman, H. 1982b. "Les questions." In C. Lefebvre, H. Magloire-Holly and N. Piou (eds), 204-233. Koopman, H. 1986. "The genesis of Haitian: Implications of a comparison of some features of the syntax of Haitian, French and West African languages." In P.C. Muysken and N. Smith (eds), 231-258. Koopman, H. and Lefebvre, C. 1981. "Haitian Creole pu." In P.C. Muysken (ed.), 201-221. Koopman, H. and Lefebvre, C. 1982. "PU : marqueur de mode, préposition et complémenteur." In C. Lefebvre, H. Magloire-Holly and N. Piou (eds), 64-91. Kouwenberg, S. 1992. "From OV to VO. Linguistic negotiation in the development of Berbice Dutch Creole." Lingua 88 (3-4): 263-299. Kouwenberg, S. 1994. A Grammar of Berbice Dutch Creole [Mouton Grammar Library 12]. Berlin/New York: Mouton de Gruyter. Kouwenberg, S. and Murray, E. 1994. Papiamentu. Languages of the World/Materials 83. Mümchen/Newcastle: Lincom Europa.

324

ISSUES IN THE STUDY OF PIDGIN AND CREOLE LANGUAGES

Kuo, C.H. 1997. "A corpus-based approach to linguistic analysis: Applications and features." Studies in English Literature and Linguistics 23 (June): 113-125. Labov, W. 1971. "On the adequacy of natural languages: The development of tense." Unpublished Ms. Labov, W. 1972a. Sociolinguistic Patterns. Philadelphia, PA: University of Pennsylvania Press. Labov, W. 1972b. Language in the Inner City: Studies in the Black English Vernacular. Philadelphia, PA: University of Pennsylvania Press. Labov, W. 1975. What is a Linguistic Fact? Lisse: The Peter De Ridder Press. Lappin, S. 1984. "Predication and raising." In C. Jones and P. Sells (eds), 236-252. Larson, R.K. and Lefebvre, C. 1991. "Predicate clefting in Haitian Creole." In Proceedings of the North East Linguistic Society, T. Sherer (ed.), 247-262. Université du Québec à Montréal: GLSA. Law, P. 1991. "Verb movement, expletive replacement, and head-government." The Linguistic Review 8: 253-285. Law, P. 1992. "What is a null subject language?" Travaux de recherche sur le créole haïtien 11: 41-58 [Université du Québec à Montréal]. Law, P. 1994a. "Une note sur la construction existentielle en Haïtien." Revue québécoise de linguistique 23 (1): 139-167. Law, P. 1994b. "Resumptives in Haitian and relexification." Paper presented at The MIT Symposium on the role of relexification in creole genesis: The case of Haitian Creole. Law, P. and Lefebvre, C. 1995. "On the relationship between event determiners and predicate cleft in the Kwa languages: The case of Fongbe." Les propriétés lexicales du créole haïtien et du fongbe. Special issue of the Linguistique Africaine 14: 7-47. Lefebvre, C. 1976. "Linguistic survey of Cuzco Quechua: Sampling procedures and data collection." Anthropological Linguistics 18 (7): 328-339. Lefebvre, C. 1982. "L'expansion d'une catégorie grammaticale : Le déterminant la In C. Lefebvre, H. Magloire-Holly and N. Piou (eds), 21-63. Lefebvre, C. 1984. "Grammaires en contact : définition et perspectives de recherche." Revue québécoise de linguistique 14 (1): 11-49. Lefebvre, C. 1986. "Relexification in creole genesis revisited: The case of Haitian Creole. In P.C. Muysken and N. Smith (eds), 279-301. Lefebvre, C. 1989. "Instrumental take-serial constructions in Haitian and in Fon." The Canadian Journal of Linguistics 34 (3): 319-337. Lefebvre, C. 1990. "On the interpretation of predicate cleft." Linguistic Review 6 (3): 169-194. Lefebvre, C. 1991a. "Take serial verb constructions in Fon." In C. Lefebvre (ed.), 37-78. Lefebvre, C. 1991b. "La distribution du déterminant et des complémenteurs en créole haïtien." In A. Kihm (ed.), 21-47.

REFERENCES

325

Lefebvre, C. 1992. "AGR in languages without person and number agreement: The case of the clausal determiner in Haitian and Fon." Functional Categories. Special issue of The Canadian Journal of Linguistics 37 (2): 137-156. Lefebvre, C. 1993a. "The role of relexification in creole genesis: The case of functional categories." Travaux de recherche sur le créole haïtien 14: 23-52 [Université du Québec à Montréal]. Lefebvre, C. 1993b. 'The role of relexification and syntactic reanalysis in Haitian Creole: Methodological aspects of a research program." In S.S. Mufwene (ed.), 254-279. Lefebvre, C. 1994a. "Functional categories and relexification." In C. Lefebvre and J.S. Lumsden (eds) (1994a), 74 pages. Lefebvre, C. 1994b. "On spelling out E." Travaux de recherche sur le créole haïtien 23: 1-33 [Université du Québec à Montréal]. Lefebvre, C. 1994c. "New facts from Fongbe on the double object constructions." Lingua 94 (2-3): 69-163. Lefebvre, C. 1996a. "The tense, mood, and aspect system of Haitian Creole and the problem of transmission of grammar in creole genesis." Journal of Pidgin and Creole Languages 11 (2): 231-311. Lefebvre, C. 1996b. 'The functional category AGR and creole genesis." In H. Wekker (ed.), 153-195. Lefebvre, C. 1997. "Relexification in creole genesis: The case of demonstrative terms in Haitian Creole." Journal of Pidgin and Creole Languages 12 (2): 181-201. Lefebvre, C. 1998a. Creole Genesis and the Acquisition of Grammar: The Case of Haitian Creole [Cambridge Studies in Linguistics 88]. Cambridge: Cambridge University Press. Lefebvre, C. 1998b. "Multifunctionality and variation among grammars: The case of the determiner in Haitian and in Fongbe." Journal of Pidgin and Creole Languages 13(1): 93-150. Lefebvre, C. 1999a. "Substratum semantics in the verbal lexicon of Haitian creole." Studies in Language 23 (1): 61-103. Lefebvre, C. 1999b. "On the empirical reliability of some Haitian data." Journal of Pidgin and Creole Languages 14 (2): 379-384. Lefebvre, C. 2000a. "What do creole studies have to offer to mainstream linguistics?" Journal of Pidgin and Creole Languages 15(1): 127-153. Lefebvre, C. 2000b. "On data." Journal of Pidgin and Creole Languages. 15 (2): 313-337. Lefebvre, C. 2001a. 'The interplay of relexification and levelling in creole genesis and development." Linguistics 39 (2): 371-408. Lefebvre, C. 2001b. "Multifunctionality and the concept of lexical entry." Journal of Pidgin and Creole Languages 16 (1): 107-145. Lefebvre, C. 2001c. "On the semantic opacity of creole languages." Journal of Pidgin and Creole Languages 16 (2): 321-353. Lefebvre, C. 2001d. "What you see is not always what you get: Apparent simplicity and hidden complexity in creole languages. Linguistic Typology 5 (2/3): 186-213.

326

ISSUES IN THE STUDY OF PIDGIN AND CREOLE LANGUAGES

Lefebvre, C. 2002. "The field of pidgin and creole linguistics at the turn of the millenium: The problem of the genesis and development of PCs." In G. Gilbert (ed.), 247-287. Lefebvre, C. 2003a. 'The genesis of pidgin and creole languages: A state of the art. " In C. Jourdan and K. Tuite (eds). Lefebvre, C. 2003b. 'The emergence of productive morphology in Haitian Creole." In G. Booij and J. van Marie (eds). Lefebvre, C. In press "Coordinating constructions in Fongbe with reference to Haitian Creole." In Coordinating Constructions. Typological Studies in Language, M. Haspelmath (ed.). John Benjamins Publishing Company. Lefebvre, C. (ed.). 1982. La syntaxe comparée du français standard et populaire : Approches formelle et fonctionnelle. Montreal: Office de la langue française. Lefebvre, C. (ed.). 1991. Serial Verbs: Grammatical, Comparative and Cognitive Approaches [Studies in the Sciences of Language Series 8]. Amsterdam/Philadelphia: John Benjamins Publishing Company. Lefebvre, C. (ed.). 1995. Research report prepared for FC AR on the project L'Organisation des lexiques et des entrées lexicales [4 volumes], Université du Québec à Montréal, Montreal. Lefebvre, C. and Brousseau, A.-M. 2002. The Structure of Fongbe [Mouton Grammar Library 25]. Berlin/New York: Mouton de Gruyter. Lefebvre, C. and Fournier, R. 1976. "Le corpus du créole haïtien à Montréal." Unpublished corpus, Université du Québec à Montréal, Montreal. Lefebvre, C. and Kaye, J. (eds). 1986. Études syntaxiques, morphologiques et phonologiques. Research report prepared for FCAR and PAFAC on the project Fon-Créole Haïtien. Université du Québec à Montréal, Montreal. Lefebvre, C. and Lumsden, J.S. 1989. "Les langues créoles et la théorie linguistique." La créolisation. Theme issue. Revue canadienne de linguistique/The Canadian Journal of Linguistics 34 (3): 249-272. Lefebvre, C. and Lumsden, J.S. 1992. "On word order in relexification." Travaux de recherche sur le créole haïtien 10: 1-22 [Université du Québec à Montréal]. Lefebvre, C. and Lumsden, J.S. 1994a. "Le rôle central de la relexification dans la genèse des langues créoles." Plurilinguismes 8 (December): 47-93. Lefebvre, C. and Lumsden, J.S. 1994b. "Relexification in creole genesis." In C. Lefebvre and J.S. Lumsden (eds) (1994a), 28 pages. Lefebvre, C. and Lumsden, J.S. 1994c. "Des différences entre le créole haïtien et le français." Espace créole 8: 189-209. Lefebvre, C. and Lumsden, J.S. (eds). 1994a. MIT Symposium on the Central Role of Relexification in Creole Genesis: The Case of Haitian Creole. Research report prepared for SSHRCC on the project La genèse du créole haïtien : un cas particulier d'investigation sur la forme de la grammaire universelle. Université du Québec à Montréal, Montreal.

REFERENCES

327

Lefebvre, C. and Lumsden, J.S. (eds). 1994b. Études syntaxiques. Research report prepared for SSHRCC and PAFAC on the project La Genèse du créole haïtien : Un cas particulier d'investigation sur la forme de la grammaire universelle [3 volumes], Université du Québec à Montréal, Montreal. Lefebvre, C , Magloire-Holly, H. and Piou, N. (eds). 1982. Syntaxe de l'haïtien. Ann Arbor, MI: Karoma. Lefebvre, C. and Massam, D. 1988. "Haitian Creole syntax: A case for DET as head." Journal of Pidgin and Creole Languages 3 (2): 213-243. Lefebvre, C. and Muysken, P.C. 1988. Mixed Categories: Nominalizations in Quechua. Dordrecht: Kluwer Academic Publishers. Lefebvre, C. and Ritter, E. 1993. "Two types of predicate doubling adverbs in Haitian Creole." In F. Byrne and D. Winford (eds), 65-91. Levin, B. 1989. "Towards a lexical organization of English verbs." Unpublished paper, Northwestern University, Evanston, IL. Levin, B. and Rappaport, M. 1986. "The formation of adjectival passives." Linguistic Inquiry 17 (4): 623-661. Levinson, S.C. 1983. Pragmatics [Cambridge Textbooks in Linguistics]. Cambridge: Cambridge University Press. Lieber, R. 1980. "On the organisation of the lexicon." Unpublished Ph.D. dissertation, MIT, Cambridge, MA. Lieber, R. 1992. Deconstructing Morphology. Chicago, IL: University of Chicago Press. Lightfoot, D. 1979. Principles of Diachronic Syntax [Cambridge Studies in Linguistics 23]. Cambridge: Cambridge University Press. Lommatzsch, E. 1925. "Deiktische Elemente im Altfranzösischen." In Jahrbuch für Philologie, V. Klemperer and E. Lerch (eds), 202-244. München: Verlag der Hochschulbuchandlung Max Hueber. Longacre, R.E. 1964. Grammar Discovery Procedures: A Field Manual. The Hague: Mouton de Gruyter. Lord, C. 1976. "Evidence for syntactic reanalysis: From verb to complementizer in Kwa." In Papers from the parasession on diachronic syntax, S.B. Steever, CA. Walker and S.S. Mufwene (eds), 179-191. Chicago, IL: University of Chicago Press. Lumsden, J.S. 1989. "On the distribution of determiners in Haitian Creole." Revue québécoise de linguistique 18 (2): 64-93. Lumsden, J.S. 1990. "The bi-clausal structure of Haitian clefts." Linguistics 28 (4): 741-760. Lumsden, J.S. 1991. "On the acquisition of nominal structures in the genesis of Haitian Creole." Travaux de recherche sur le créole haïtien 4: 37-61 [Université du Québec à Montréal]. Lumsden, J.S. 1993. "Expletives in double object constructions in Haitian Creole." In F. Byrne and D. Winford (eds), 269-284. Lumsden, J.S. 1994. "Possession: Substratum semantics in Haitian Creole." Journal of Pidgin and Creole Languages 9 (1): 25-51. Lumsden, J.S. 1995a. "Aspect and lexical semantic representations in Haitian Creole." The Linguistic Review 12 (2): 123-142.

328

ISSUES IN THE STUDY OF PIDGIN AND CREOLE LANGUAGES

Lumsden, J.S. 1995b. "On the complex nature of grammatical simplification in pidgins and early creole languages." Paper presented at the Conference of the Society for Pidgin and Creole Linguistics (LSA), New Orleans. Lumsden, J.S. 1996. "On the acquisition of nominal structures in the genesis of Haitian Creole." In H. Wekker (ed.), 184-205. Lumsden, J.S. 1999a. "Language acquisition and creolization." In M.A.F. DeGraff (ed.), 129-158. Lumsden, J.S. 1999b. 'The role of relexificaiton in creole genesis." Journal of Pidgin and Creole Languages 14 (2): 225-258. Lumsden, J.S. and Lefebvre, C. 1994. "The genesis of Haitian Creole." In C. Lefebvre and J.S. Lumsden (eds) (1994a), 15 pages. Maclaran, R. 1980. "On two asymmetrical uses of the demonstrative determiners in English." Linguistics 18 (9-10): 803-820. Magloire-Holly, H. 1982. "Les modaux : Auxiliaires ou verbes?" In C. Lefebvre, H. Magloire-Holly and N. Piou (eds), 92-120. Manfredi, V. 1993. "Verb focus in the typology of Kwa/Kru and Haitian." In F. Byrne and D. Winford (eds), 3-51. Mann, C.C. 1993. "Polysemic functionality of prepositions in pidgins and creoles: The case of 'Fò' in Anglo-Nigerian pidgin." In F. Byrne and J. Holm (eds), 57-67. Marantz, A. 1982. "Re reduplication." Linguistic Inquiry 13 (3): 435-483. Marantz, A. 1997. "No escape from syntax: Don't try morphological analysis in the privacy of your own lexicon." Unpublished Ms., MIT, Cambridge, MA. Massam, D. 1989. "Predicate argument structure in Haitian Creole." Revue québécoise de linguistique 18 (2): 95-130. Massam, D. 1990. "Cognate objects as thematic objects." Canadian Journal of Linguistics 35 (2): 161-190. Maurer, P. 1988. "Les réitérations et reduplications lexicalisées du Papiamento : Influence du substrat Africain?" In N. Boretzky, W. Enninger, and T. Stolz (eds), 95-118. McDaniel, D., McKee, C. and Cairns, H.S. (eds). 1996. Methods for Assessing Children's Syntax. Cambridge, MA: The MIT Press. McWhorter, J.H. 1997a. "It happened at Cormantin: Locating the origin of the Atlantic English-based creoles." Journal of Pidgin and Creole Languages 12 (1): 59-102. McWhorter, J.H. 1997b. "Lost in transmission: A case for the independent emergence of the copula in Atlantic creoles." In A.K. Spears and D. Winford (eds), 241-264. McWhorter, J.H. 1998a. "Identifying the creole prototype: Vindicating a typological class." Language 74 (4): 788-818. McWhorter, J.H. 1998b. The Word on the Street: Fact and Fable about American English. New York, NY: Plenum Press. McWhorter, J.H. 2001. "The world's simplest grammars are creole grammars." Linguistic Typology 5 (2): 125-166. Mensah, E.N.A. 1983. "Le krobou." In G. Hérault (ed.), 425-463. Merlan, F.C. 1994. A Grammar of Wardaman [Mouton Grammar Library 11]. Berlin/New York: Mouton de Gruyter.

REFERENCES

329

Meyer-Lübke, W. 1909. "Le français au Canada." Bulletin du français parlé au Canada 8: 121-129. Migeod, F.W.H. 1911. The Languages of West Africa. London: Kegan Paul. Migge, B.M. 1998a. "Substrate influence in creole formation: The origin of give-type serial verb constructions in the Surinamese Plantation Creole." Journal of Pidgin and Creole Languages 13 (2): 215-265. Migge, B.M. 1998b. "Substrate influence in the formation of the Surinamese Plantation Creole: A consideration of sociohistorical data and linguistic data from Ndyuka and Gbe." Unpublished Ph.D. dissertation, Department of Linguistics, Ohio State University, Columbus, OH. Milner, J.-C. 1978. De la syntaxe à l'interprétation. Paris: Éditions du Seuil. Mintz, S.W. 1971. "The socio-historical background to pidginization and creolization." In D. Hymes (ed.), 4 8 1 - 9 8 . Möhlig, W.J.G. 1983. "Mbugu." In Lexicon der Afrikanistik, H. Jungraithmayr and W.J.G. Möhlig, (eds), 158-159. Berlin: Dietrich Reimer. Moore, K.E., Peterson, D.A. and Wentum, C. (eds). 1994. Proceedings of the Twentieth Annual Meeting of the Berkeley Linguistics Society. Berkeley, CA: Berkeley Linguistics Society. Moreau, M.-L. 1971. "L'homme que je crois qui est venu : que, qui relatifs et conjonctions." Langue française 11: 77-90. Morrill, C.-H. 1997. "Language, culture, and society in the Central African Republic: The emergence and development of Sango." Unpublished Ph.D. dissertation, Indiana University, Bloomington, IN. Mougeon, R. and Beniak, É. (eds). 1994. Les origines du français québécois. Sainte-Foy: Presses de l'Université Laval. Mous, M. 1994. "Ma'a or Mbugu." In P. Bakker and M. Mous (eds), 175-200. Mous, M. 1995. "Language intertwining." Paper presented at the Amsterdam Creole Workshop: Creole Genesis and Language Contact, Amsterdam. Mous, M. In press. "Ma'a as an ethno-register of Mbugu." Sprache und Geschichte in Afrika 15. Mufwene, S.S. 1986. 'The universalist and substrate hypotheses complement one another." In P.C. Muysken and N. Smith (eds), 129-162. Mufwene, S.S. 1989. "Some explanations that strike me as incomplete." Journal of Pidgin and Creole Languages 4 (1): 117-128. Mufwene, S.S. 1990. "Transfer and the substrate hypothesis in creolistics." Studies in Second Language Acquisition 12 (1): 1-23. Mufwene, S.S. 1991. "Pidgins, creoles, typology, and markedness." In F. Byrne and T. Huebner (eds), 123-144. Mufwene, S.S. 1992. "Africanisms in Gullah: A re-examination of the issues." In Old English and New: Studies in Language and Linguistics in Honor of Frederic G. Cassidy, J.H. Hall, N. Doane and D. Ringler (eds), 156-182. New York, NY: Garland. Mufwene, S.S. 1993. "Scope of negation and focus in Gullah." In F. Byrne and D. Winford (eds), 95-116. Mufwene, S.S. 1994a. "Restructuring, feature selection, and markedness: From Kimanyanga to Kituba." In K.E. Moore, D.A. Peterson and C. Wentum (eds), 18-21.

330

ISSUES IN THE STUDY OF PIDGIN AND CREOLE LANGUAGES

Mufwene, S.S. 1994b. "On decreolization: The case of Gullah." In Language and the Social Construction of Identity in Creole Situations, M. Morgan (ed.), 63-99. Los Angeles, CA: Center for Afro-American Studies, University of California at Los Angeles. Mufwene, S.S. 1994c. "Genèse de population et genèse de langue." Créolistique et grammaire genérative 8 (December): 95-113. Mufwene, S.S. 1996a. "Creolization and grammaticalization: What creolistics could contribute to grammaticalization." In P. Baker and A. Syea (eds), 5-28. Mufwene, S.S. 1996b. "Creole genesis: A population genetics perspective." In Caribbean Language Issues: Old and New, P. Christie (ed.), 163-196. Kingston, Jamaica: University of the West Indies Press. Mufwene, S.S. 1997a. "Jargons, pidgins, creoles, and koines: What are they?" In A.K. Spears and D. Winford (eds), 35-70. Mufwene, S.S. 1997b. "Kituba." In S.G. Thomason (ed.), 173-208. Mufwene, S.S. 2000. "Creolization is a social, not a structural, process." In I. Neumann-Holzschuh and E.W. Schneider (eds), 65-84. Mufwene, S.S. (ed.). 1993. Africanisms in Afro-American Language Varieties. Athens, GA/London: The University of Georgia Press. Mühlhäusler, P. 1980. "Structural expansion and the process of creolisation." In A. Valdman and A.R. Highfield (eds), 19-56. Mühlhäusler, P. 1985. "The lexical system of Tok Pisin." In Handbook of Tok Pisin (New Guinea Pidgin) [Pacific Linguistics Series C. 70], S.A. Wurm and P. Mühlhäusler (eds), 423-440. Canberra: The Australian National University. Mühlhäusler, P. 1986a. Pidgin and Creole Linguistics [Language in Society II]. Oxford: Basil Blackwell. Mühlhäusler, P. 1986b. "Bonnet blanc and blanc bonnet: Adjective-noun order, substratum and language universals." In P.C. Muysken and N. Smith (eds), 41-55. Mühlhäusler, P. 1997. Pidgin and Creole Linguistics [Westminster Creolistics Series 3]. London: University of Westminster Press. Mulder, J. 1996. "Methodology and description." Linguistique 32 (1): 17-34. Muysken, P.C. 1981a. "Half-way between Quechua and Spanish: The case for relexification." In Historicity and Variation in Creole Studies, A.R. Highfield and A. Valdman (eds), 52-79. Ann Arbor, MI: Karoma. Muysken, P.C. 1981b. "Creole tense/mood/aspect systems: The unmarked case?" In P.C. Muysken (ed.), 181-200. Muysken, P.C. 1987. "Prepositions and postpositions in Saramaccan." In Studies in Saramaccan Language Structure [Caribbean Culture Studies 2], M.C. Alleyne (ed.), 89-102. Amsterdam: Universiteit van Amsterdam. Muysken, P.C. 1988a. "Lexical restructuring in creole genesis." In N. Boretzky, W. Enninger and T. Stolz (eds). Muysken, P.C. 1988b. "Are creoles a special type of language?" In Linguistic Theory: Extensions and Implications [Linguistics: The Cambridge Survey 2], F.J. Newmeyer (ed.), 285-301. Cambridge: Cambridge University Press.

REFERENCES

331

Muysken, P.C. 1988c. "Media Lengua and linguistic theory." The Canadian Journal of Linguistics 33 (4): 409-422. Muysken, P.C. 1988d. "Parameters for serial verbs." In Niger-Congo Syntax and Semantics 7, K.A. Demuth and V. Manfredi (eds). Boston, MA: Boston University. Muysken, P.C. 1994a. 'The first seven years, fat or lean?" Journal of Pidgin and CreoleLanguages9 (1): 103-107. Muysken, P.C. 1994b. "Saramaccan and Haitian: A comparison." Journal of Pidgin and Creole Languages 9 (2): 305-314. Muysken, P.C. 1995. "Studying variation in older texts: Negerhollands." Journal of Pidgin and Creole Languages 10 (2): 335-347. Muysken, P.C. (ed.). 1981. Generative Studies on Creole Languages [Studies in Generative Grammar 6]. Dordrecht: Foris Publications. Muysken, P.C. and Smith, N. 1990. "Question words in pidgin and creole languages." Linguistics 28 (4): 883-903. Muysken, P.C. and Smith, N. 1995. "Reflexives." In J. Arends, P.C. Muysken and N. Smith (eds), 271-287. Muysken, P.C. and Smith, N. (eds). 1986. Substrata Versus Universals in Creole Genesis. Papers from the Amsterdam creole workshop, April 1985 [Creole Language Library 1]. Amsterdam/Philadelphia: John Benjamins Publishing Company. Nagao, M. 1994. "A methodology for the construction of a terminology dictionary." In Computational approaches to the lexicon, B.T.S. Atkins and A. Zampolli (eds), 397-441. Oxford: Oxford University Press. Naro, A.J. 1978. "A study of the origins of pidginisation." Language 54 (2): 314-347. Neumann-Holzschuh, I. and Schneider, E.W. (eds). 2000. Degrees of Restructuring in Creole Languages [Creole Language Library 22]. Amsterdam/Philadelphia: John Benjamins Publishing Company. Newmeyer, F.J. (ed.). 1988. Language: The Socio-Cultural Context [Linguistics: The Cambridge Survey 4]. Cambridge: Cambridge University Press. Nida, E. 1948. "The identification of morphemes." Language 24: 414-441. Nyrop, K. 1936 Formation des mots [Grammaire historique de la langue française III]. Copenhagen: Gyldendalske Boghandel, Nordisk Forlag. Orjala, P.R. 1970. "A dialect survey of Haitian Creole." Unpublished Ph.D. dissertation, Hartford Seminary Foundation, Kansas City, MO. Owens, J. 1997. "Arabic-based pidgins and creoles." In S.G. Thomason (ed.), 125-172. Pagliuca, W. (ed.). 1994. Perspectives on Grammaticalization [Current Issues in Linguistic Theory 109]. Amsterdam/Philadelphia: John Benjamins Publishing Company. Papen, R. 1978. The French-Based Creoles of the Indian Ocean: An Analysis and Comparison. Ann Arbor/London: University Microfilms International. Papen, R. 1988. "Convergence et divergence en métif." Paper presented at the Département de linguistique, Université du Québec à Montréal, Montreal.

332

ISSUES IN THE STUDY OF PIDGIN AND CREOLE LANGUAGES

Partee, B. 1973. 'The syntax and semantics of quotation." In A Festschrift for Morris Halle, S. Anderson and P. Kiparsky (eds), 410-418. New York, NY: Holt, Rinehart & Winston. Pasch, H. 1997. "Sango." In S.G. Thomason (ed.), 209-270. Patrick, P.L. 1999. Urban Jamaican Creole. Variation in the Mesolect [Varieties of English Around the World G 17]. Amsterdam/Philadelphia: John Benjamins Publishing Company. Payne, T.E. 1997. Describing Morphosyntax: A Guide for Field Linguists. Cambridge: Cambridge University Press. Peleman, L.F. 1978. Diksyonnè kréol-fransè. Port-au-Prince: Bon Nouvèl. Peleman, L.F. 1986. fi Diksyonnè Kréyòl-Nélandè. Port-au-Prince: Bon Nouvèl. Perry, T. and Delpit, L. (eds). 1998. The Real Ebonics Debate: Power, Language, and the Education of African-American Children. Boston, MA: Beacon Press. Piou, N. 1982a. "Le clivage du prédicat." In C. Lefebvre, H. Magloire-Holly and N. Piou (eds), 122-151. Piou, N. 1982b. "Le redoublement verbal." In C. Lefebvre, H. Magloire-Holly and N. Piou (eds), 152-166. Piwek, P.L.A. and Cremers, A.H.M. 1996. "Dutch and English demonstratives: A comparison." Language Sciences 18 (3-4): 835-851. Plag, I. 1993. Sentential Complementation in Sranan. On the Formation of an English-Based Creole Language [Linguistische Arbeiten 300]. Tübingen: Max Niemeyer. Plag, I. 1994a. "Creolization and language change: A comparison." In D. Adone and I. Plag (eds), 3-21. Plag, I. 1994b. "On the diachrony of creole complementizers: The development of Sranan taki and dati." Amsterdam Creole Studies 11: 67-14. Plag, I. 1995. 'The emergence of taki as a complementizer in Sranan: On substrate, universals, and gradual creolization." In J. Arends (ed.), 113-148. Plag, I. 1999. Morphological Productivity. Structural Constraints in English Derivation. Berlin/New York: Mouton de Gruyter. Plag, I. 2000. "Review of Claire Lefebvre Creole Genesis and the Acquisition of Grammar." Journal of Linguistics 36 (1): 176-180. Plag, I. 2001. "The nature of derivational morphology in creoles and noncreoles." Journal of Pidgin and Creole Languages 16 (1): 153-160. Plag, I. 2002. "On the role of grammaticalization in creolization. A reassessment." In G. Gilbert (ed.), 229-246. Platt, M.M. 1992. "Les constructions comparatives en fongbè, un cas de sérialisation." Travaux de recherche sur le créole haïtien 7: 91-98 [Université du Québec à Montréal]. Poirier, C. 1994. "La langue parlée en Nouvelle-France : vers une convergence des explications." In R. Mougeon and É. Beniak (eds), 237-274. Pollock, J.-Y. 1989. "Verb movement, universal grammar, and the structure of IP." Linguistic Inquiry 20 (3): 365-424.

REFERENCES

333

Pompilus, P. 1976. Contribution à l'étude comparée du créole et du français à partir du créole haïtien : Morphologie et syntaxe. Port-au-Prince: Éditions Caraïbes. Poplack, S. and Tagliamonte, S. 1996. "Nothing in context: Past temporal reference in Nigerian Pidgin English." In P. Baker and A. Syea (eds), 71-95. Popper, K.R. 1972. Objective Knowledge. Oxford: Oxford University Press. Post, M. 1995. "Fa d'Ambu." In J. Arends, P. C. Muysken and N. Smith (eds), 191-204. Pranka, P.M. 1983. "Syntax and word formation." Unpublished Ph.D. dissertation, MIT, Cambridge, MA. Press, I.J. 1986. A Grammar of Modern Breton [Mouton Grammar Library 2]. Berlin/New York: Mouton de Gruyter. Reinecke, J.E. 1964. "Trade jargons and creole dialects as marginal languages." In Language in Culture and Society, D. Hymes (ed.), 534-546. New York/Evanston/London: Harper & Row Publishers. Rey, A. (ed.). 1992. Dictionnaire historique de la langue française [2 volumes]. Paris: Le Robert. Rice, K.D. 1989. A Grammar of Slave [Mouton Grammar Library 5]. Berlin/New York: Mouton de Gruyter. Rickford, J.R. 1987. Dimensions of a Creole Continuum: History, Texts and Linguistic Analysis of Guyanese Creole. Stanford: Stanford University Press. Rickford, J.R. 1997. "Unequal partnership: Sociolinguistics and the African American speech community." Language in Society 26 (2): 161-197. Rickford, J.R. and Rickford, A.E. 1995. "Dialect readers revisited." Linguistics and Education 7 (2): 107-128. Rickford, J.R. and Rickford, R.J. To appear. Spoken Soul: The Story of Black English. New York, NY: John Wiley & Sons. Ritter, E. 1991. "Déplacement de NP en haïtien : oui ou non." In A. Kihm (ed.), 65-86. Ritter, E. 1992. "Cross-linguistic evidence for number phrase." Functional Categories. Special issue of The Canadian Journal of Linguistics 37 (2): 197-219. Rivard, A. 1914. Etudes sur les parlers de France au Canada. Québec: Garneau. Roberge, Y. 1990. The Syntactic Recoverability of Null Arguments. Kingston: McGill-Queen's University Press. Roberts, S.J. 1998. 'The role of diffusion in the genesis of Hawaiian Creole." Language 74 (1): 1-39. Roberts, SJ. 1999. "Response to Goodman." Language 75 (2): 349-351. Robertson, I. 1993. "The Ijo element in Berbice Dutch and the pidginisation/creolisation process." In S.S. Mufwene (ed.), 296-316. Romain, J.-B. 1959. Quelques mœurs et coutumes des paysans haïtiens. Travaux pratiques d'ethnographie sur la, région de Milot à l'usage des étudiants. Port-au-Prince: Imprimerie de l'État. Romaine, S. 1988. Pidgin and Creole Languages. London/New York: Longman.

334

ISSUES IN THE STUDY OF PIDGIN AND CREOLE LANGUAGES

Romaine, S. 1992. Language, Education and Development. Urban and Rural Tok Pisin in Papua New Guinea [Oxford Studies in Language Contact]. New York, NY: Clarendon Press/Oxford University Press. Romaine, S. 1999. 'The grammaticalization of the proximative in Tok Pisin." Language 75 (2): 322-346. Rongier, J. 1979. "Description de l'éwé [vol. 1]." Unpublished Ph.D. dissertation, Université Paris VII (Denis Diderot), Paris. Rongier, J. 1995. Dictionnaire français-éwé suivi d'un index français éwéfrançais. Paris: Karthala. Rosset, T. 1911. Les origines de la prononciation moderne étudiées au 17ième siècle. Paris: A. Colin. Ruhl, C. 1989. On Monosemy: A Study in Linguistic Semantics. Albany, NY: State University of New York Press. Safir, K. 1982. "Syntactic Chains and the Definiteness Effect." Unpublished Ph.D. dissertation, MIT, Cambridge, MA. Samarin, W.J. 1967. Field Linguistics: A Guide to Linguistic Field Work. New York, NY: Holt, Rinehart and Winston. Samarin, W.J. 1994. "The dynamics of morphotactic change in Sango." In K.E. Moore, D.A. Peterson and C. Wentum (eds), 125-138. Samarin, W.J. 1997. "The creolization of pidgin morphophonology." In A.K. Spears and D. Winford (eds), 175-216. Sandeman, B. 1998. "Substrate influence in the TMA system of Tayo." Unpublished Ph.D. dissertation, University of New England, Australia. Sankoff, G. 1971. The Sociolinguistic Situation of Tok Pisin in Papua New Guinea. Port Moresby, Papua New Guinea: New Guinea Research Unit. Sankoff, G. 1980. "Variability and explanation in language and culture: Cliticization in New Guinea Tok Pisin." In G. Sankoff (ed.), 257-270. Sankoff, G. 1984. "Substrate and universals in the Tok Pisin verb phrase. Meaning, form and use in context: Linguistic applications." In Georgetown University Round Table on Language and Linguistics, D. Schiffrin (ed.), 104-119. Washington, DC: Georgetown University Press. Sankoff, G. 1990. "The grammaticalization of tense and aspect in Tok Pisin and Sranan." Language Variation and Change 2 (3): 295-312. Sankoff, G. 1991. "Using the future to explain the past." In F. Byrne and T. Huebner (eds), 61-74. Sankoff, G. 1993. "Focus in Tok Pisin." In F. Byrne and D. Winford (eds.), 117-140. Sankoff, G. (ed.). 1980. The Social Life of Language. Philadelphia, PA: University of Pennsylvania Press. Sankoff, G. and Brown, P. 1980. 'The origins of syntax in discource: A case study of Tok Pisin relatives." In G. Sankoff (ed.), 211-255. Sankoff, G. and Laberge, S. 1973. "On the acquisition of native speakers by a language." Kivung 6 (1): 32-47. Sankoff, G. and Laberge, S. 1980. "On the acquisition of native speakers by a language." In G. Sankoff (ed.), 195-209. Schuchardt, H. 1909. "Die Lingua Franca." Zeitschrift für romanische Philologie 33.

REFERENCES

335

Schuchardt, H. 1979. The Ethnography of Variation: Selected Writings on Pidgins and Creoles. Ann Arbor, MI: Karoma. Schumann, J.H. 1978. The Pidginization Process: A Model for Second Language Acquisition. Rowley, MA: Newbury House. Schutze, C.T. 1996. The Empirical Base of Linguistics: Grammaticality Judgments and Linguistic Methodology. Chicago, IL: University of Chicago Press. Schwegler, A. 2002. "Creolistics in Latin America: Past, present, and future." In G. Gilbert (ed.), 121-171. Sebba, M. 1981. "Derivational regularities in a creole lexicon: The case of Sranan." Linguistics 19 (1): 101-117. Sebba, M. 1987. The Syntax of Serial Verbs. An Investigation into Serialization in Sranan and other Languages [Creole Language Library 2]. Amsterdam/Philadelphia: John Benjamins Publishing Company. Segurola, R.P.B. 1963. Dictionnaire fon-fronçais. Cotonou, Benin: Procure de l'archidiocèse. Senft, G. 1986. Kilivila. The Language of the Trobriand Islanders [Mouton Grammar Library 3]. Berlin/New York: Mouton de Gruyter. Seuren, P.A.M. 1995. "Notes on the history and syntax of Mauritian Creole." Linguistics 33 (3): 531-577. Seuren, P.A.M. and Mufwene, S.S. (eds). 1990. Issues in Creole Linguistics. Special issue of Linguistics 28 (4). Seuren, P.A.M. and Wekker, H. 1986. "Semantic transparency as a factor in creole genesis." In P.C. Muysken and N. Smith (eds), 57-71. Siegel, J. 1985. "Koines and Koineization." Language in Society 14 (3): 357-378. Siegel, J. 1987. Language Contact in a Plantation Environment: A Sociolinguistic History of Fiji. Cambridge: Cambridge University Press. Siegel, J. 1995. "Koine formation and creole genesis." Paper presented at the Amsterdam Creole Workshop: Creole Genesis and Language Contact, Amsterdam. Siegel, J. 1997. "Mixing, levelling and pidgin/creole development." In A.K. Spears and D. Winford (eds), 111-149. Siegel, J. 1999a. "Transfer constraints and substrate influence in Melanesian pidgin." Journal of Pidgin Creole Languages 19 (1): 1-44. Siegel, J. 1999b. "Creoles and minority dialects in education: An overview." Journal of Multilingual and Multicultural Development 20 (6): 508-531. Siegel, J. 2000. Processes of Language Contact: Studies from Australia and the South Pacific [Collection champs linguistiques]. Montreal: Fides. Siegel, J. 2002. "Mixing, levelling and pidgin/creole development." In Pidgins and Creoles: Structure and Status, A.K. Spears and D. Winford (eds), 11-149. Amsterdam: John Benjamins Publishing Company. Singler, J.V. 1988. "The homogeneity of the substrate as a factor in pidgin/creole genesis." Language 64 (1): 27-51. Singler, J.V. 1993a. "African influence upon Afro-American language varieties: A consideration of sociohistorical factors." In S.S. Mufwene (ed.), 235-253.

336

ISSUES IN THE STUDY OF PIDGIN AND CREOLE LANGUAGES

Singler, J.V. 1993b. "The African presence in Caribbean French colonies in the seventeenth century: Documentary evidence." Travaux de recherche sur le créole haïtien 16-17: 1-236 [Université du Québec à Montréal]. Singler, J.V. 1996. "Theories of creole genesis, sociohistorical considerations, and the evaluation of evidence: The case of Haitian creole and the relexification hypothesis." Journal of Pidgin and Creole Languages 11 (2): 185-231. Smith, N. 1987. 'The genesis of the creole languages of Surinam." Unpublished Ph.D. dissertation, Universiteit van Amsterdam, Amsterdam. Smith, N. 1995. "Voodoo chile: Why did they do it differently?" Paper presented at the Amsterdam workshop on creole genesis and language contact. Smith, N., Robertson, I. and Williamson K. 1987. "The Ijo element in Berbice Dutch." Language in Society 16 (1): 49-89. Southworth, F.C. 1971. "Detecting prior creolization: An analysis of the historical origins of Marathi." In D. Hymes (ed.), 255-275. Spagnolo, L.M. 1933. Ban Grammar. Verona: Missioni Africane. Spears, A.K. 1990. "Tense, mood, and aspect in the Haitian creole preverbal marker system." In Pidgin and creole tense-mood-aspect systems [Creole Language Library 6], J.V. Singler (ed.), 119-142. Amsterdam/Philadelphia: John Benjamins Publishing Company. Spears, A.K. and Winford, D. (eds). 1997. The Structure and Status of Pidgins and Creoles [Creole Language Library 19]. Amsterdam/Philadelphia: John Benjamins Publishing Company. Sproat, R.W. 1985. "On Deriving the Lexicon." Unpublished Ph.D. dissertation, MIT, Cambridge, MA. Sterlin, M.-D. 1988. "Les différentes caractéristiques de pou en créole haïtien." Travaux de recherche sur le créole haïtien 3: 1-34 [Université du Québec à Montréal]. Sterlin, M.-D. 1989. "Les caractéristiques de pou : Un modal en position de complémenteur." Le créole haïtien. Special issue of Revue québécoise de linguistique 18 (2): 131-147. Stewart, W.A. 1962. "Creole languages in the Caribbean." In Study of the Role of Second Languages in Asia, Africa and Latin America, F.A. Rice (ed.), 34-53. Washington, DC: Center for Applied Linguistics for Modern Language Association. Stewart, W.A. 1967. "Sociolinguistic factors in the history of American Negro dialects." Florida FL Reporter 5. Syea, A. 1997. "Copula, wh-trace, and the ECP in Mauritian Creole." Linguistics 35 (1): 25-56. Sylvain, S. 1936. Le créole haïtien : morphologie et syntaxe. Wetteren, Belgium: Imprimerie De Meester/Port-au-Prince: By the author. Tardif, O. 2000. "Multifonctionalité et sous-spécification : Le cas de ɖò en fongbe." Unpublished MA thesis, Université du Québec à Montréal, Montreal. Tardif, O. no date. 'The morphology of French in the 17th century." Unpublished Ms., Université du Québec à Montréal, Montreal.

REFERENCES

337

Thiele, J. 1987. La formation des mots en français moderne. Montreal: Les Presses de l'Université de Montréal. Thoiron, P., Iwaz, J. and Zaouche, N. 1997. "The methodology of analysis of results of surveys on the implantation of the official terminology." Terminologies nouvelles 16 (June): 48-54. Thomason, S.G. 1997a. "A typology of contact languages." In A.K. Spears and D. Winford (eds), 71-88. Thomason, S.G. 1997b. "Introduction." In S.G. Thomason (ed.), 1-7. Thomason, S.G. 2001. Language Contact. Edinburgh: Edinburgh University Press. Thomason, S.G. (ed.). 1997. Contact Languages: A Wider Perspective [Creole Language Library 17]. Amsterdam/Philadelphia: John Benjamins Publishing Company. Thomason, S.G. and Kaufman, T. 1991. Language Contact, Creolization, and Genetic Linguistics [2nd edition]. Berkeley/Los Angeles/Oxford: University of California Press. Thornell, C. 1995. "Established French loanwords in Sango: A pilot study." Working Papers Lund University 44: 159-177. Thornell, C. 1997. "The Sango language and its lexicon (Senda-yanga tisango)." Unpublished Ph.D. dissertation, Lunds University, Ann Arbor, MI. Tinelli, H. 1970. "Generative phonology of Haitian Creole." Unpublished Ph.D. dissertation, University of Michigan, Ann Arbor, MI. Tinelli, H. 1981. Creole Phonology. The Hague: Mouton de Gruyter. Traugott, E.C. 1999. "Review of Caribbean language issues: Old and new." Journal of Pidgin and Creole Languages 14 (2): 208-213. Traugott, E.C. and Heine, B. 1991. "Introduction." In E.C. Traugott and B. Heine (eds), vol. 1, 1-14. Traugott, E.C. and Heine, B. (eds). 1991. Approaches to Grammaticalization [2 volumes] [Typological Studies in Language 19]. Amsterdam/Philadelphia: John Benjamins Publishing Company. Travis, L. 1992. "Inner aspect and the structure of VP." Cahiers de linguistique de l'UQAM 1: 132-149. Travis, L. 2000. "L-syntax/S-syntax boundary: Evidence from Austronesian." In Formal issues in Austronesian linguistics, P. Ileana and V. Phillips (eds), 167-194. Dordrecht: Kluwer Academic Publishers. Tremblay, M. 1991. "Possession and datives: Binary branching from lexicon to syntax." Unpublished Ph.D. dissertation, McGill University, Montreal. Tresbarats, C. 1983. "L'abidji". In G. Hérault (ed.), 41-65. Trudgill, P. 1986. Dialects in Contact [Language in Society 10]. Oxford: Basil Blackwell. Valdman, A. 1970. Basic Course in Haitian Creole. The Hague: Mouton de Gruyter. Valdman, A. 1978. Le créole : Structure, statut et origine. Paris: Klincksieck. Valdman, A. 1980. "Creolization and second language acquisition." In A. Valdman and A.R. Highfield (eds), 297-311.

338

ISSUES IN THE STUDY OF PIDGIN AND CREOLE LANGUAGES

Valdman, A. 1993. "Matrice et diffusion culturelle dans le développement des créoles français." Paper presented at the Université du Québec à Montréal, November. Valdman, A. 1996. A Learner's Dictionary of Haitian Creole. Bloomington, IN: Indiana University, Creole Institute. Valdman, A. (ed.). 1977. Pidgin and Creole Linguistics. Bloomington/London: Indiana University Press. Valdman, A. (ed.). 1979. Le français hors de France [Créoles et français régionaux]. Paris: Honoré Champion. Valdman, A. and Highfield, A.R. (eds). 1980. Theoretical Orientations in Creole Studies. New York, NY: Academic Press. Valdman, A. et al. 1981. Haitian Creole-English-French Dictionary [2 volumes]. Bloomington, IN: Indiana University, Creole Institute. Valois, D. 1991. "The internal syntax of DP." Unpublished Ph.D. dissertation, University of California, Los Angeles, CA. Van Dyk, P. [ca. 1765]. nieuwe en nooit bevoorens geziene onderwyzinge in het Basiert Engels, of Neeger Engels, zoo als het zelve in de Hollansze Colonien gebruikt word. Amsterdam: Erven de Weduwe Jacobus van Egmont. Van Name, A. 1869-1870. "Contribution to creole grammar." Transactions of the American Philological Association 1869-70: 123-167. Vaux, B. and Cooper, J. 1999. Introduction to Linguistic Field Methods. Harvard: Lincom Europa. Védrine, E.W. 1992. Dictionary of Haitian Creole Verbs (with phrases and idioms). Cambridge, MA: Soup to Nuts Publishers. Veenstra, T. 1992. "Double objects in Haitian Creole." In Études syntaxiques, vol. 1, C. Lefebvre and J.S. Lumsden (eds), 170-193. Research report prepared for SSHRCC, FCAR and PAFAC on the project La genèse du créole haïtien : un cas particulier d'investigation sur la forme de la grammaire universelle [2 volumes]. Université du Québec à Montréal, Montreal. Veenstra, T. 1996a. Serial Verbs in Saramaccan: Predication and Creole Genesis. Dordrecht: Holland Institute of Generative Linguistics. Veenstra, T. 1996b. "Grammaticalized verbs in Saramaccan." In P. Baker and A. Syea (eds), 95-112. Vernet, A. et al. 1976. Ti diksyonnè kreyòl-franse: dictionnaire élémentaire créole haïtien-français. Port-au-Prince: Éditions Caraïbes. Véronique, D. 1994. "Naturalistic adult acquisition of French as L2 and Frenchbased creole genesis compared: Insights into creolization and language change?" In D. Adone and I. Plag (eds), 117-138. Vilsaint, F. 1992. Diksyonè Anglè Kreyol English Kreyòl Dictionary. Temple Terrace, FL: Educa Vision. Vinet, M.-T. 1991. "Observations théoriques sur les explétifs: le cas du créole haïtien." Revue québécoise de linguistique 20 (1): 195-214. Voorhoeve, J. 1973. "Historical and linguistic evidence in favor of the relexification theory in the formation of creoles." Language in Society 2: 133-145.

REFERENCES

339

Voorhoeve, J. 1981. "Multifunctionality as a derivational problem." In P. Muysken (ed.), 25-34. Walker, J.A. and Samarin, W.J. 1997. "Sango phonology." In Phonologies of Asia and Africa (including the Caucasus), A.S. Kaye (ed.), 861-880. Winona Lake, IN: Eisenbrauns. Wallace, É. 1995. "Le groupe nominal en éwé." In C. Lefebvre (ed.), vol. 2, 15 pages. Washabaugh, W. 1975. "On the development of complementisers in creolisation." Working Papers on Language Universals 17 (June): 109-140. Weinreich, U. 1953. Languages in Contact. The Hague: Mouton de Gruyter. Weinreich, U. 1984. "Lexicographic definition in description semantics." Alpha 28: 103-118. Wekker, H. 1982. 'The transparency principle in second language acquisition and creolization." Paper presented at the Fourth Biennial Conference of the Society for Caribbean Linguistics, Paramaribo, 1-6 Sept. 1982. Wekker, H. (ed.). 1996. Creole Languages and Language Acquisition [Trends in Linguistics, Studies and Monograph 86]. Berlin/New York: Mouton de Gruyter. Westerman, D. 1930. A Study of the Ewe Language. London: Oxford University Press. Westerman, D. and Bryan, M.A. 1970. The Languages of West Africa [New edition]. Folkestone/Daw sons: The Gresham Press. Whinnom, K. 1956. Spanish Contact Vernaculars in the Philippine Islands. Hong Kong: Hong Kong University Press. Whinnom, K. 1965. 'The origin of the European-based creoles and pidgins." Orbis 14 (2): 509-527. Whinnom, K. 1971. "Linguistic hybridisation and the 'special case' of pidgins and creoles." In D. Hymes (ed.), 91-115. Whinnom, K. 1977. "Lingua franca: Historical problems." In A. Valdman (ed.), 295-310. Winford, D. 1993. Predication in Caribbean English Creoles [Creole Language Library 10]. Amsterdam/Philadelphia: John Benjamins Publishing Company. Winford, D. 1996. "Common ground and creole TMA." Journal of Pidgin and Creole Languages 11 (1): 71-84. Winford, D. 1997. "Creole studies and sociolinguistics." Journal of Pidgin and CreoleLanguages12 (2): 303-318. Winford, D. 2000. ""Intermediate" creole and degrees of change in creole formation: The case of Bajan." In I. Neumann-Holzschuh and E.W. Schneider (eds), 215-246. Wolf, L. 1991. "Le Langage de la cour et le français canadien". In Français du Canada-Français de France [Proceedings of the 2 e Colloque international de Cognac, France, September 27-30, 1988], B. Horiot (ed.), 115-123. Tübingen: Niemeyer. Wolf, L. 1994. "Postface." In R. Mougeon and E. Beniak (eds), 325-332. Woolford, E. 1983. "Introduction: The social context of creolisation." In E. Woolford and W. Washabaugh (eds), 1-9.

340

ISSUES IN THE STUDY OF PIDGIN AND CREOLE LANGUAGES

Woolford, E. and Washabaugh, W. (eds). 1983. The Social Context of Creolisation. Ann Arbor, MI: Karoma Publishers. Wurm, S.A., Mühlhäusler, P. and Tryon, D.T. (eds). 1996. Atlas of Languages of Intercultural Communication in the Pacific, Asia, and the Americas. Berlin/New York: Mouton de Gruyter. Yvon, H. 1951."Cil et cist, articles démonstratifs." Romania 72: 145-181. Zribi-Hertz, A. and Bingaba-Ngaima, S. 1994. "Les pronoms topicaux du sango et la théorie du liage." Recherches linguistiques de Vincennes 23: 99-135.

APPENDIX 1

A RESEARCH PROGRAMME ON PC GENESIS FOR THE 21ST CENTURY Since my research interests focus on the processes involved in the genesis of pidgins and creoles (henceforth PCs), I will limit my discussion of further research on PCs to this topic. Setting the problem of creole genesis within the framework of the processes otherwise known to play a role in language genesis and change in general distinguishes the discussion of the problem from other current approaches (e.g. substratist, superstratist and universalist approaches). In this view, language universals are part of PCs just as they are part of other natural languages and the language-specific features of PCs are predicted by the nature of the main process at work in their formation: relexification. Given the process of relexification, the specific features of PCs are predicted to reproduce those of their substratum languages. This prediction is borne out (see chapter 3). The other two processes, reanalysis and dialect levelling, are fed by the first (see chapter 9). I believe that more studies of this type will be needed in the future. Since the methodology required to test the role of these processes in PC genesis is essentially based on comparative work, I shall begin with a discussion of the types of comparative work that will be needed for future research. Comparative work The results of the comparison of Solomons Pidgin with its source languages (Keesing 1988) and of Haitian with its source languages (Lefebvre 1998a, and the references therein) are similar: both Solomons Pidgin and Haitian Creole reproduce the semantic and syntactic properties of their respective substratum languages. This result is striking and leads to the conclusion that the problem of PC genesis should no longer be addressed on the basis of data drawn from PCs alone. I therefore urge scholars interested in PC genesis to get involved in serious comparative work of a PC and its source languages. For this to be possible, historical studies like that done by Singler (1996) for the UQAM Haitian project, that will establish who were the right people, in the right place, at the right time, are required. So, I fully agree with Arends (2002) that historical studies will be very welcome if scholars in the field perform comparative work using appropriate language samples.

342

ISSUES IN THE STUDY OF PIDGIN AND CREOLE LANGUAGES

I believe that the field would also benefit from another type of comparative work: comparison of different creoles formed on the basis of the same pool of substratum languages but different lexifier languages (e.g. Haitian and Saramaccan or Sranan). This type of comparative work should lead to a better understanding of the contribution of superstratum languages to PCs in general. Furthermore, it will allow us to evaluate the limits imposed by the superstratum data on the formation of PC lexicons and eventually to explain some of the differences between creoles formed from the same substratum languages but different superstratum languages. As was mentioned in section 2.4.2, relabelling is constrained by what the superstratum language has to offer as a suitable form to relabel a lexical entry copied from the substratum languages. For example, the substratum final determiner could be relabelled as la because French had a suitable form là to relabel the copied lexical entry. On the assumption that English does not offer a comparable form, the copied lexical entry could not have been relabelled in English based creoles such as Sranan or Saramaccan. This would explain why in Haitian the determiner appears to have been created through relexification whereas in Sranan it appears to have been created through reanalysis, as per the account in Bruyn (1995). Comparison of a given PC's substratum languages and the PC itself is also needed. If relexification is indeed the main process at work in PC genesis, and given that relexification is an individual activity, speakers of different substratum languages will tend to reproduce the idiosyncrasies of their respective lexicons in creating a creole. Comparing the lexicons of the substratum languages with each other will reveal differences among them. When compared with the PC's lexicon, these differences will reveal early cases of dialect levelling, that is, cases of levelling taking place at the time the PC is formed. There is yet another type of comparative work which I see as important for a better understanding of the general issue of PC genesis: comparison of PCs that have been formed from the same pool of substratum languages and the same superstratum language (e.g. Haitian and Martinican Creole). What would this type of comparison reveal? Again, it would shed light on another aspect of dialect levelling. The relative impact of relexification and reanalysis in PC genesis and development There is no agreement among scholars as to what types of lexical items may undergo relexification. [For example, Muysken (1988a) has claimed that functional categories do not undergo relexification and Muysken (1988c) affirms that affixes cannot be relexified. The position I take with respect to this issue is the following: all lexical entries can be copied but only those which have some semantic content can be assigned a new label on the basis of a superstratum

A RESEARCH PROGRAMME ON PC GENESIS FOR THE 21ST CENTURY

343

phonetic sequence. This is in line with Muysken's assertion that relexification (relabelling in my terms) is semantically driven. Hence, in my view, all lexical entries with some semantic content may be relabelled including functional categories and derivational affixes. Lexical entries that have no semantic content at all (e.g. Case markers, operators, etc.) cannot be relabelled and are thus assigned a phonologically null form when relabelling takes place. Practically speaking, this means that when required in a PC utterance, these lexical entries are not pronounced (on phonologically null forms in relexified lexicons, see also Lefebvre and Lumsden 1994b).] This disagreement is correlated to the relative importance of relexification and reanalysis in the formation of PCs. Researchers who allow for functional categories and derivational affixes to undergo relexification—provided that they have some semantic content—find fewer cases of reanalysis in the data than those who do not. Note, however, that in the long run the result is the same. The tension in the field concerning the relative impact of relexification and reanalysis in PC genesis and development must, however, be resolved. It must be possible to list the types of arguments that would lead to a resolution of this issue and to assess the data in light of these arguments. In this respect, the paper by Plag (2002) presents a list of principled and independent criteria to identify true cases of reanalysis. Work carried out on early PC data such as Sankoff s on Tok Pisin or more recent work on Nigerian Pidgin English (e.g. Poplack and Tagliamonte 1996) is also greatly needed. In my view, however, this type of work would benefit a great deal from a detailed documentation of the subsystems of the substratum lexicons. It has recently been observed that developing PCs tend to manifest features of their substratum languages even when the latter cease to be spoken in the PC community (e.g. Mufwene 1990; Mühlhäusler 1986a, 1986b; Sankoff 1991, etc.). In Lefebvre and Lumsden (1992, 1994b), it was suggested that this striking phenomenon can be explained by the relationship between reanalysis and relexification. As these authors show, a functional category lexical entry that has been assigned a null form at relabelling may be signalled by a periphrastic expression. Through reanalysis, that periphrastic expression may later become the phonological representation of a lexical entry initially assigned a null form. When the null form becomes overt, it has the properties of the substratum lexical entry. The emergence of the Haitian definite future marker is discussed along these lines in Lefebvre (1996b). Much more work needs to be done on this problem before an understanding of exactly what is going on can be reached. The question to be addressed here is the following: how can the first generation of native speakers of a particular PC deduce the properties of the grammatical system of their parents on the basis of the data they are presented with? This brings me to a related and still controversial issue: the respective

344

ISSUES IN THE STUDY OF PIDGIN AND CREOLE LANGUAGES

contributions of adults and children to PC genesis and development. Having completed an extensive comparison of Haitian and its source languages (Lefebvre 1998a), I have no doubt that the input of adult speakers of the substratum languages is extremely important. Again, my findings are similar to those of Keesing (1988) for Solomons pidgin. I assume that the contribution of the first generation of native speakers to the development of Haitian is similar to that reported by Sankoff for Tok Pisin. That is, in a dramatic case of reanalysis such as the one discussed by Sankoff (1991: 73) the "native speakers are building on in carrying forth more sweeping changes than their parents were able to do." But not all lexical entries are created so dramatically. In fact, most are not. Furthermore, Sankoff's work has shown us that in several cases (e.g. iabracketing, cliticisation, etc.), fluent adult non-native speakers of Tok Pisin are participating in changing the language along with the first generation of native speakers. I have no reason to believe that the same situation did not prevail at the time the substratum languages of Haitian ceased to be spoken in Haiti. An area of research that is most likely to reveal the important contribution made to the PC by the first generation of native speakers is the development of morphophonemics. Dialect levelling It is only in recent literature that dialect levelling has been proposed to play a role in the development of PCs (e.g. Harris 1991; Mühlhäusler 1980). This process has seldom been documented. Thus, it is a prime topic for further research in the field. It has already been pointed out that the criteria governing the selection among competing forms require further investigation (e.g. Mufwene 1986, 1990). Siegel (2002) presents a preliminary discussion of this point (see also Lefebvre 1998a).

APPENDIX 2

THE COMPOSITION OF THE HAITIAN LEXICON This appendix provides an overview of the types of Haitian lexical entries with respect to origin that I found in the course of my research on Haitian Creole. The content of this appendix draws from section 2.2 of Lefebvre (2002). Seven types have been identified: retentions from the West African substratum languages, lexical items acquired from French, borrowings from neighbouring languages, lexical entries produced by relexification, lexical entries that have acquired a phonological representation through reanalysis, lexical entries that have been subjected to levelling, and finally, innovations. First, there are Haitian lexical entries that, in addition to having the semantic and syntactic properties of the corresponding lexical entries in West African languages, also have a phonological representation derived from these languages. I consider lexical entries with such properties to be cases of retention from the West African substratum languages spoken in Haiti at the time the creole was formed. Hilaire (1993) lists about 350 such words still in use in modern Haitian. These words are mainly from Ewe, Fongbe, Mandingo and Kikongo. Hilaire's list includes nouns referring to religious beliefs, objects, persons or activities, body parts, food, various objects, animals, plants, kinship relationships, etc.; some verbs and a few exclamations have also been retained. Second, there are Haitian lexical entries which, in addition to having a phonological representation derived from French also have the properties of the corresponding French lexical entries. These lexical entries have no equivalents in the substratum languages. I consider such lexical entries to be cases of acquisition from French. Note that this type of lexical entry is found only among major category lexical items. No French functional category lexical entries were acquired by the creators of Haitian for there are no such lexical entries in modern Haitian. Third, the Haitian lexicon also contains words of other origins. Amerindian languages spoken on the island at the time of the early Haitian colony contributed some 200 nouns designating local objects such as plants and place names (see Hilaire 1992). A few words of Spanish origin (see Hilaire 1992) and a few borrowings from English are also part of the Haitian lexicon (see Valdman et al. 1981). These constitute cases of lexical borrowing.

346

ISSUES IN THE STUDY OF PIDGIN AND CREOLE LANGUAGES

Fourth, the bulk of the Haitian lexicon comprises lexical entries that have the semantic and syntactic properties of substratum language lexical entries but phonological representations derived from French, thus, lexical entries that have been produced by the process of relexification. This type of lexical entry is found among major category lexical entries, including derivational affixes, and among functional category lexical entries. Fifth, as in other languages, there are some Haitian lexical entries which can be argued to have acquired a label through the process of reanalysis. The following cases of reanalysis have been found in the course of this research. The first case involves the reanalysis of the verb bay 'to give' as a preposition ba(y) meaning 'for' (Déchaine 1988). This type of reanalysis involves a change in one of the syntactic values that defines verbs; i.e. [-N, +v]→[-N, -V]. It is visible in the syntax: as a second verb in a series, bay cannot be extracted with its object, while as a preposition it can. This kind of reanalysis is well documented in several languages of the world, including West African languages (see Lord 1973). As we saw in section 3.3.4.3, there is no Haitian lexical entry having the properties of the substratum postnominal item known as the indefinite determiner. Modern Haitian has a so-called prenommai indefinite determiner yon 'a'. In Lefebvre (1998a: 87-89), based on work by Brousseau (in preparation), it is proposed that this Haitian form evolved from the reanalysis of the Haitian prenommai numeral younn 'one' (< French une 'one (feminine)') as yon 'a'. The reanalysis is visible in the resyllabification of younn as yon. Sankoff (1991) reports a dramatic case of reanalysis that took place in Tok Pisin. This case consists in the reanalysis of the sentence-initial adverb baimbai 'after' as an aspectual marker occurring between the subject and the verbs with the semantic properties of the corresponding substratum lexical entry. In Lefebvre (1996b) the Haitian definite future marker is hypothesised to have evolved from the presentential adverb apre 'after'. Apre apparently moved to the position between the subject and the verb, where it was then reanalysed as an irrealis marker, more specifically a definite future marker, manifesting the semantic properties of the corresponding marker in West African languages. The above examples exhaust the cases of reanalysis that I have found in my detailed survey of the Haitian Creole lexicon. All other potential cases of reanalysis can be better analysed as cases of relexification when Haitian is compared with its contributing languages. For example, Koopman and Lefebvre (1981) suggested a reanalysis account of the emergence of the complementiser pou in Haitian, which was argued to have become a complementiser through the reanalysis of the preposition pou 'for' introducing purposive complements and of the irrealis marker pou. While this analysis may be convincing when the Haitian data are analysed in isolation, a comparison with data from the

THE COMPOSITION OF THE HAITIAN LEXICON

347

substratum languages casts doubt on this account. For example, in Fongbe, the irrealis marker ní is also used as a complementiser, as is the preposition nú introducing purposive complements. Hence, the additional functions of the Haitian irrealis marker pou and preposition pou have their source in substratum language lexical entries (for further discussion, see Lefebvre 1998a). This finding is congruent with Bruyn's (1996), who shows that some grammaticalisation chains were adopted integrally from Gbe into Sranan. The findings presented by Lefebvre (1998a) based on Haitian are congruent with those of Bruyn (1996) and Plag (2002) based on Sranan: convincing cases of reanalysis in the development of these creoles are hard to find. Sixth, in my in-depth survey of the Haitian lexicon, I also found several cases of Haitian lexical entries which have a phonological representation derived from French and the properties of one (but not all) of the substratum languages. I consider that such lexical entries result from relexification followed by dialect levelling. Recall that dialect levelling operates on the output of the relexification of various substratum lexicons. Cases of dialect levelling include the plural marker occurring in nominal structure, the Genitive case marker (see Lumsden and Lefebvre 1994), some TMA markers (see Lefebvre 1998a), reflexive forms (see Carden and Stewart 1988; Lefebvre 1998a), and the nominalising affix (see Lefebvre 1998a). Finally, there are Haitian lexical entries whose properties correspond to none of the contributing languages. I consider these cases, of which I found four, as innovations from within the creole. The Haitian existential verb is one such case. Indeed, the properties of the verb gen 'there is' do not correspond to those of any of Haitian's source languages (see also Koopman 1986). This lexical entry therefore consitutes a case of development from within the creole, a conclusion which is in line with that in DeGraff (1992b). Likewise, the properties of the Haitian lexical entry se cannot be associated with any source language lexical entry. In agreement with DeGraff (1992b), I consider se to constitute another independent development from within the creole. In my analysis of Haitian derivational morphology, the verbalising suffix -e and the adverbial suffix -man appear to be cases of independent development inhanced by French. The above classification exhausts the types of lexical entries that I have found in my survey of the Haitian lexicon. The three processes that have been proposed to play a role in creole genesis and development in general can thus be argued to have played a major role in the genesis and development of Haitian in particular. According to my evaluation, the bulk of the Haitian lexicon was created by relexification, sometimes followed by reanalysis and/or dialect levelling.

INDEX OF AUTHORS A. Abney, S. 172 Adam, L. 10, 37, 127 Adamson, L. 304 Adone,D. 23, 125, 131 Agbidinoukoun, C.C. 91, 151 Allen, M.R. 262, 267 Alleyne, M.C. 9-10, 14, 16-18, 37, 42, 126-129, 135, 235, 301 Amberber, M. 174 Andersen, R.W. 9, 16, 42, 129, 209, 235, 301 Anonymous 64-65, 94, 96, 100-101, 173, 211,254 Ans, A.-M. 91, 149-150, 250, 253, 260, 286, 303 Aoun, J. 143, 145 Arends, J. 131, 137, 156, 180, 341 Aronoff, M. 262, 267 Avolonto, A. 112, 145 Awoyale, Y. 61,246 B. Baker, M.C. 112-113 Baker, P. 9, 18, 23, 26, 125, 131, 137, 144, 156, 212 Bakker, P. 23, 25, 39, 135, 235, 298, 304 Bámgbósé, A. 26 Bastide, R. 30, 48 Bauer, L. 262 Baugh, J. 138 Beniak, É. 239 Bennis, H. 110 Bentolila, A. 95, 97, 126, 199, 211 Bickerton, D. 9, 14, 16, 19-21, 26, 32, 45, 111, 120-123, 125-126, 128, 135136, 140, 199, 237, 305 Bingaba-Ngaima, S. 132 Bloomfield, L. 12, 142 Bolé-Richard, R. 255 Botha, R.P. 143, 145 Bouchard, D. 103, 166, 167 Bouquiaux, L. 142 Braun, M. 298, 304, 308 Bresnan, J. 174

Brousseau, A.-M. 29, 31, 40, 51-55, 5859, 61-66, 68, 74, 80-82, 86, 91-92, 94, 111, 125-126, 139, 150, 155, 165, 167, 170, 181-182, 188, 192, 208, 214, 216, 218, 221, 225-226, 228, 233, 242, 245-246, 251, 259260, 263-264, 266, 268-272, 277, 291, 293-294, 296-298, 304, 346 Brown, P. 20, 156 Brunot, F. 247,251 Bruyn, A. 27, 131, 136-137, 156, 160, 175, 179, 342, 347 Bryan, M.A. 297, 299 Buba, M. 255 Burling, R. 142 Byrne, F. 134, 155, 159, 178 C. Cadely, J.-R. 111,227 Campbell, L. 142 Capo, H.B.C. 1, 244, 298 Carden, G. 31, 43, 62, 74, 144, 246-247, 249, 347 Cassidy, F. 298 Cattell, R. 81 Chansou, M. 140 Chaudenson, R. 9, 16-18, 33, 46, 128129, 135, 144, 192, 235, 239, 247, 251,265 Chomsky, N. 19, 59, 110-111, 168, 174, 185, 196, 220, 248 Clark, E. 165 Clark, H. 165 Claudi, U. 175 Clements, G.N. 56 Clements, J.C. 131-132 Collins, C. 102, 117, 119, 222 Comrie, B. 37, 56 Cooper, J. 142 Corbin, D. 262 Corne, C. 9, 125, 132, 144 Cowart, W. 153 Cowper, E.A. 166-167 Cremers, A.H.M. 255 Crowley, T. 262 Curtin, P.D. 21

350

ISSUES TN THE STUDY OF PIDGIN AND CREOLE LANGUAGES

D. Da Cruz, M. 108 Damoiseau, R. 97, 199, 265 De Guiraudon, T.G. 139 DeCamp, D. 298 Déchaîne, R.-M. 61, 97, 113, 158-159, 165, 199, 246-247, 346 Dees A. 251 DeGraff, M.A.F. 5, 58, 77, 91, 105-107, 111-112, 114, 126-127, 130, 137, 140, 144, 147, 149, 182, 193-194, 196-197, 250, 255, 260, 265, 273, 275-276, 279, 285-286, 288-289, 291, 306-307, 347 Dejean, Y. 144, 147 Delafosse, M. 241 Delpit, L. 138 Denis, M.-J. 259, 267, 273, 292 Déprez, V. 58, 105, 107, 111, 114-115, 130, 139, 147, 153, 197, 212 Di Sciullo, A.-M. 84, 260, 267 Diessel, H. 255 Dijkhoff, M. 298, 304 Dillard, J.L. 32, 45, 237 Domingue, N. 23, 27, 43, 231 Donaldson, B.C. 142 Donohue, M. 143, 165 Doran, M. 147 Drapeau, L. 147 Dubois, J. 294-295 Dumais, D. 72-73, 75-76, 197-198 Duponchel, L. 256 Durie, M. 232 E. Escure, G. 131, 138, 155 Étienne, G. 94, 145, 250, 253, 256 F. Faine, J. 74, 91, 149, 239, 250, 274, 286, 288-291 Faltz, L.M. 61, 246 Fattier, D. 153 Fauchois, A. 170, 268 Férère, G.A. 94, 145, 250, 253, 255-256 Ferguson, C.A. 12 Ferraz, L. 10, 128 Filipovich, S. 86, 126, 170, 182, 188189, 192-193, 195, 260, 263-264, 266-269, 291, 293 Fillmore, C.J. 255 Foley, W.A. 8, 13, 26, 47, 129, 238 Fournier, R. 91, 147-149 Frake, C.O. 15 Freeman, B.C. 263-264, 268, 270, 274275, 278-279, 286-288, 290-291, 293

G. Gacic, M. 140 Gambhir, S.K. 236 Germain, R. 289, 291 Ghomeshi, J. 166 Gilbert, G.G. 7, 35, 125, 138 Gilles, R. 104, 163 Gilman, C. 156 Givón, T. 112 Göbl-Galdi, L. 12 Goddard, Y. 142 Goldsmith, J. 100 Goodman, M.F. 9, 11-12, 14, 23, 34, 47, 54, 74, 91, 94, 122, 127-128, 133, 137, 145, 149, 154, 237, 241-242, 246-247, 250, 252 Gougenheim, G. 242 Grenoble, L. 255 Grevisse, M. 5 6 , 9 7 , 2 5 1 Grimes, B.F. 2 Grimshaw, J. 15, 81 Guiraud,P. 251 Gundel, J.K. 255 H. Haase, A. 56, 261, 296 Hagège, C. 10, 56, 126, 217 Hale,K. 167, 171 Hall, R.A. 9, 12, 15, 23, 122, 128, 136, 146, 158, 170, 250, 252, 260, 268, 289-290, 298, 303 Hancock, I.F. 6, 10, 14-15, 47, 125, 128 Harris, J.W. 27, 136, 231, 236, 344 Haspelmath, M. 26 Hay, J. 262 Hazaël-Massieux, G. 128, 137 Hazoumê, M.L. 56, 59, 107, 245 Hedberg, N. 255 Heine, B. 26, 156, 175, 177, 208 Hérault, G. 255-256 Herskovits, M.J. 30, 48 Hesseling, D.C. 9-10, 126, 128, 217 Highfield, A.R. 27 Hilaire, J. 287, 308, 345 Hinskens, F. 155 Hirschbühler, P. 100 Holm, J. 10, 17-18, 127, 134 Hopper, P.J. 2, 26, 122, 128, 175 Huber, M. 132, 137 Hulk, A. 110,212 Hull, A. 239 Hünnemeyer, F. 175 Huttar, G.L. 11, 37, 47, 127-128, 155 Hyams, N.M. 111 Hyman, L.M. 56 Hymes, D. 5, 8, 10, 14-15, 23, 47, 128129

INDEX OF AUTHORS I. Isquerdo, A.N. 144 Issoufi, O.A. 144

J. Jackendoff, R. 168 Jaeggli, O.A. 110,212 Jaggar, P.J. 255 Jensen, B. 159, 187 Jespersen, O. 12 Johns, A. 166 Johnson, K. 113-114,212 Joseph, F. 91, 94, 145, 149, 250, 253, 255-256 Joseph, J.S. 108, 157, 163 Jourdan, C. 7, 23, 27, 125, 127-128, 138, 231,233 Julliand, A. 295 Juneau, M. 239-240, 295 K. Kaufman, T. 8, 10, 16, 23, 42, 47, 128129, 205, 209, 235-236, 238, 301 Kay, P. 6, 15 Kaye, J. 29-30, 42, 48, 208 Kayne, R.S. 83, 98, 100, 102, 165 Keesing, R.M. 11, 16, 18, 20, 27-28, 3435, 47, 123, 126-128, 137, 180, 214, 215-216, 341, 344 Keyser, J. 167, 171 Kibrik, A.K. 142 Kihm, A. 131-132 Kinyalolo, K.K.W. 54, 57, 59, 61, 78, 98, 102, 173, 180, 222-223, 245-246, 248 Kiple, K. 21 Koanting, E. 155 Koopman, H. 21, 25-27, 58, 61-63, 69, 71-73, 77-79, 83, 95, 99-100, 102104, 116, 120, 126, 130, 146, 163, 187, 197, 204, 213, 346-347 Kouwenberg, S. 142, 226, 298, 304 Kuo, C.H. 144 L. Laberge, S. 14-15, 20, 27, 40, 97, 156, 177 Labov, W. 14, 23, 140, 142, 148, 152153, 232, 237 Laguerrè, J. 263, 270, 291, 293 Lappin, S. 198 Larson, R.K 117, 202-203, 224 Law, P. 58, 76, 102, 108, 110-111, 117, 147, 197

351

Lefebvre, C. 2, 4, 6-9, 11-13, 15-18, 2021, 23, 25-35, 37-43, 45-55, 57-65, 67-74, 76-79, 81-100, 102-105, 107108, 110-119, 121-122, 125-129, 135-139, 144-145, 147-150, 152, 154-155, 163-167, 170-172, 178, 180-188, 192, 196, 199, 201-203, 205, 207-214, 216-217, 219-227, 231-248, 250-255, 257, 259-260, 263-264, 267-269, 273, 279, 285, 290-297, 299-301, 303, 308, 341, 343-347 Levin, B. 141, 269, 294 Levinson, S.C. 255 Lieber, R. 84, 170, 269 Lightfoot, D. 9, 26, 43, 175 Lommatzsch, E. 251 Longacre, R.E. 142 Lord, C. 26, 98, 165, 173, 346 Lumsden, J.S. 8, 11-12, 15-17, 20, 29-31, 33-34, 39-40, 42-43, 46-49, 53, 60, 68, 83, 91, 94, 97, 102, 105, 128130, 136-137, 144-145, 147, 149150, 154, 164, 167, 184-185, 197199, 208-209, 214, 217, 219-223, 226, 232-238, 240-242, 247, 251, 301, 343 M. Maclaran, R. 255 Magloire-Holly, H. 95 Manfredi, V. 61, 102, 246-247 Mann, C.C. 155 Marantz, A. 171,296 Massam, D. 58, 69, 76, 81, 91, 108, 148149, 166, 172, 197-198, 200, 204 Maurer, P. 298 McDaniel, D. 142 McWhorter, J.H. 4, 35, 125, 127-128, 133, 137-138, 181-182, 204-205, 207, 217, 220, 225-229, 305, 308 Meister, A. 81 Mel, G. 256 Mensah, E.N.A. 254 Merlan, F.C. 143 Meyer-Lübke, W. 239 Migeod, F.W.H. 241 Migge, B.M. 34, 137 Milner, J.-C. 98 Mintz, S.W. 10, 128 Möhlig, W.J.G. 23 Moreau, M.-L. 102 Morrill, C.-H. 132 Mougeon, R. 239 Mous, M. 2, 23, 25, 39, 41, 135, 208, 234-235

352

ISSUES IN THE STUDY OF PIDGIN AND CREOLE LANGUAGES

Mufwene, S.S. 6, 10, 15-16, 18, 27, 31, 33, 42-43, 46, 61, 125-127, 129, 132134, 136, 155, 160, 175, 217, 219, 222, 231, 235-236, 241, 248, 304, 343-344 Mühlhäusler, P. 6, 15, 26-27, 33, 46, 131, 133, 136, 155-156, 231, 305, 343-344 Mulder, J. 143 Murray, E. 298 Muysken, P.C. 2, 20-21, 23-26, 28, 32, 38-40, 59, 74, 113, 123, 125-127, 130-131, 135, 140, 144, 165, 171, 180, 187, 208, 212, 220, 226, 234235, 246-249, 298, 342 N. Nagao, M. 140 Naro, A.J. 13 Neumann-Holzschuh, I. 18 Nida, E. 166 Nikiema, E. 294 Nyrop, K. 295 O. Orjala, P.R. 153, 237, 250, 256, 271, 283, 287 Owens, J. 133 P. Pagliuca, W. 156 Papen, R. 23, 304 Parkvall, M. 132, 137 Partee, B. 174 Pasch, H. 132 Patrick, P.L. 129, 131, 138 Payne, T.E. 142 Peleman, L.F. 250, 253, 276, 287, 291 Perry, T. 138 Piou, N. 197 Piwek, P.L.A. 255 Plag, I. 23, 26, 125, 128, 136-138, 156, 160, 174, 178-179, 181, 205, 259, 262, 298, 304, 308, 347 Platt, M.M. 187 Poirier, C. 239 Pollock, J.-Y. 111, 196,212 Pompilus, P. 146, 250, 256, 287 Poplack, S. 129, 343 Popper, K.R. 143 Post, M. 304 Pranka, P.M. 121 Press, I.J. 142

289-

155186, 343,

R. Rappaport, M. 269, 294 Reh, M. 26, 175, 177, 208 Reinecke, J.E. 126 Rey, A. 295 Rice, K.D. 142 Rickford, A.E. 138 Rickford, J.R. 26, 138 Rickford, R.J. 138 Riley, M. 255 Ritter, E. 91-92, 118, 149, 172, 197 Rivard, A. 239 Roberge, Y. 110,212 Roberts, S.J. 133, 137 Robertson, I. 8, 61, 222, 248 Romain, J.-B. 237 Romaine, S. 26, 133, 136, 156, 167 Rongier, J. 254 Rosset, T. 251 Ruhl, C. 166-167 S. Safir, K. 110 Samarin, W.J. 132, 142 Sandeman, B. 126 Sankoff, G. 6, 14-15, 18, 20, 23, 26, 33, 40, 45, 97, 123, 126-129, 136, 142, 156, 175, 177, 219, 236, 343-344, 346 Schneider, E.W. 18 Schuchardt, H. 10, 12, 126, 217 Schumann, J.H. 16, 42, 129, 209, 235, 301 Schutze, C.T. 145 Schvaneveldt, R. 262 Schwegler, A. 125, 132 Sebba, M. 160, 298 Segurola, R.P.B. 64, 76, 96, 254 Senft, G. 142 Seuren, P.A.M. 4, 19, 125-126, 133, 181, 186-187, 196-197, 203-204, 304-305 Siegel, J. 7, 23, 27-28, 33, 35, 43, 45, 125, 129, 136-138, 231, 233, 235236, 344 Singler, J.V. 8-10, 21, 30, 48-49, 121, 128, 236-238, 257, 296, 301, 308, 341 Smith, N. 8, 26, 28, 59, 74, 125, 131, 135, 137, 180, 222, 246-249, 298, 304 Southworth, F.C. 15 Spagnolo, L.M. 177 Spears, A.K. 95-96, 131, 164, 211 Sproat, R.W. 121 Sterlin, M.-D. 77-79, 98-99, 147, 163 Stewart, W.A. 14, 25, 31, 43, 62, 74, 113, 144,246-247,249, 347

INDEX OF AUTHORS Syea, A. 23, 26, 131, 133, 156 Sylvain, S. 11, 37, 47, 61, 74, 91, 94, 122, 145-146, 149, 158, 163, 170, 188-189, 199, 242, 246-247, 250, 253, 266, 268, 284, 287, 289-291, 298 T. Tagliamonte, S. 129, 343 Tardif, O. 173-174, 180, 259, 295 Thiele, J. 294 Thoiron, P. 140 Thomas, J.M.C. 142 Thomason, S.G. 8, 10, 16, 23, 42, 47, 127-129, 131, 134-135, 205, 209, 235-236,238,301,305 Thorneil, C. 132 Tinelli, H. 94, 145, 153, 170, 237, 250, 252, 260, 268, 287 Traugott, E. 27 Traugott, E.C. 2, 26-27, 122, 128, 156, 175 Travis, L. 81, 155, 172 Tremblay, M. 83 Tresbarats, C. 231,236 Trudgill, P. 2, 23, 27, 43, 208 Tryon, D.T. 133 Tuite, K. 7 V. Valdman, A. 9, 13, 16, 27, 42, 52, 62, 72, 91, 94, 96, 129, 141, 145-146, 149, 153, 155, 157-159, 161-162, 164, 166-167, 183-184, 188-190, 192-195, 211, 216, 235, 239, 250, 253, 255-256, 259-260, 263-268, 270-272, 274-281, 285, 287-293, 298, 301, 303-308, 345

353

Valois, D. 172 Van Dyk, P. 179 Van Name, A. 9, 128 Van Rossem, C. 155 Vaux, B. 142 Védrine, E.W. 184, 186, 263, 269, 287 Veenstra, T. 61, 83, 155, 159-160, 166, 177, 180, 222, 248, 298, 304 Vernet, A. 287, 291 Véronique, D. 16 Vilsaint, F. 94, 250, 255, 287, 289, 291 Vinet, M.-T. 58, 105 Voorhoeve, J. 9, 11, 25, 37, 47, 127-128, 156 W. Walker, J.A. 132 Wallace, É. 241,254 Washabaugh, W. 27 Weinreich, U. 140, 235 Wekker, H. 4, 19, 125-126, 129, 181, 186-187, 196-197, 203-204, 304-305 Westerman, D. 241, 254, 297, 299 Whinnom, K. 8, 14, 23, 26, 47, 129, 232 Williams, E. 84, 260, 267 Williamson K. 8 Winford, D. 18, 125, 130-131, 133, 156, 160 Wolf, L. 239 Woolford, E. 6, 16 Y. Yvon, H. 251 Z. Zacharski, R. 255 Zribi-Hertz, A. 132

INDEX OF SUBJECTS A. anaphor 58-62, 222-223 argument structures types of 69-71 see argument alternations see verbs argument alternations 200-201 B. bare sentences temporal and aspectual interpretation of 97-98 BODY-part nouns 51-52 BODY-part reflexives 61-62 see compounds C. case case assigning properties of verbs 83 case markers within the noun phrase 94, 185 genitive case in nominal structures 5354,185 categorial underspecification 168-175 underspecification account of Fongbe <ɖò 'to say': verbe and complementiser 173175 clausal determiner 108-109 see definite determiner clause structure functionnal categories in 95-110 see cleft markers see complementisers see conjunction of clauses see evidential see negation markers see operator see tense, mood and aspect markers cleft markers 105-106 clitics see pronouns comparative constructions 187 complementisers 98-102 introducing complements of verbs of the say-class 98-100

introducing complements of verbs of the want-class 98-100 or resumptives in the context of extracted subjects 102 reanalysed from verbs 173-175 compounds 50-52 conjunction of clauses 104-105 of NPs l05 control verbs see verbs creole languages as isolating languages 4, 217-219 do creole languages form a typological class? 4, 207-229 features of 4, 8-12, 225-228 genetic affiliation of 207-229 hidden complexity of 223-224 opacity of 181-206 pidgins and creoles as PCs 5-6 similarity of 208-217 simplicity of 220-223 transparency of 181-206 typological features of 207, 225-229 creole studies and linguistic theory 3, 125-138 creolistic see creole studies D. data (on) 3, 139-154 inconsistencies in 148-153 limits of particular types of 143-148 the non-neutral character of 140-143 definite determiner in nominal structure 89-92, 148-151, 209-210 in the clause 90, 108-109 deictic terms see demonstratives demonstratives 92-94, 144, 210, 250-257 determiners definite see definite determiner demonstrative see demonstratives

356

ISSUES IN THE STUDY OF PIDGIN AND CREOLE LANGUAGES

indefinite see indefinite determiner order of 88-89 dialect levelling 231-258 cases of 241-257 definition of 27, 231 in creole genesis and development 231 double-object verbs see verbs E. event determiner see definite determiner evidentials 108 see clausal determiner existential verbs see verbs expletives see pronouns F. functional categories see cleft marker see complementisers see determiners see evidentials see focus marker see negation marker see operator see tense, mood and aspect markers future definite 96-97 indefinite 96-97 see mood G. Gbe dialect cluster 1 genesis of creole languages competing theories of nativised pidgins 14-16 reduced codes 12-14 relexification 24-26, 43-44 restructured varieties 17-21 second language acquisition 16-17 problem of the 8, 23, 46 grammaticalisation and multifunctionnality 175-178 apparent cases of grammaticalisation in creole languages 178-180 definition of 175 see reanalysis grammaticality judgments 154

H. Haiti situation at the time of creole genesis 48, 238-241 Haitian lexicon (composition of) 345-347 cases of acquisition 345 cases of borrowing 345 cases of innovation 347 cases of levelling 347 cases of relexification 346 cases of reanalysis 346 cases of retention 345 I. indefinite determiner 92 L. lexical entry concept of 165-166 see monosemy principle M. markedness marked/unmarked structures 120-121 mood markers 95-98 see evidentials see future monosemy principle 165-168 morphology 5, 84-88, 188-196, 259-308 conversion 268-272 derivational affixes 84-88 emergence of the morphological inventory of Haitian Creole 292-303 evaluating the productivity of affixes 262-292 identification of affixes in creole languages 260-262 prefixes of Haitian Creole 188, 261-268, 273 suffixes of Haitian Creole 188-196, 261272, 275-292 see compounds see Semantic Transparency Hypothesis multifunctional lexical items range 161-165 see multifunctionality multifunctionality 3, 155-180, 186 and grammaticalisation 175-178 definition of 155 establishing the multifunctionality of lexical items 157-161 relexification of multifunctional items 180 see categorial underspecification

INDEX OF SUBJECTS N. negation markers 106-107 negative quantifiers 114-116 nominal structures 88-89 bare NPs 91 conjunction of NPs 105 functional categories involved in nominal structures 88-95 word order in 88-89 see case see determiners see plural marker nouns deverbal nouns 283-285, 291, 306 see compounds see morphology null subject see parameters O. operator 102-104, 221-222 P. Parameters 110-120 double-object constructions 113-114, 212 negative quantifiers 114-116, 212 null subject parameter 110-111, 212 parameter setting in creole genesis 110, 212 serial verbs 112-113,212 verb doubling phenomena 116-120, 212 verb raising 111-112,212 plural marker 90-91, 148-151, 210, 241242 possessive forms possessive adjectives 54, 153-154, 210 see pronouns predicate cleft construction 116-120, 145, 202-203,213,223-224 pronouns expletives 57-58 logophoric 56-57 personal 52-54, 216, 241- 242 possessive 54-56 pronominal clitics 111, 218-219 reciprocal see anaphor reflexive see anaphor resumptive 198 see resumptives see wh-words

357

R. raising 197 see verbs reanalysis 2, 33, 43-46, 342, 344 and related phenomena 26 cases of 45 definition of 26 reciprocal see anaphor reflexives anaphor 58-61, 245-250 body-part 60-61, 245-250 relabelling definition of 39-40, 44, 60 see relexification relexification and dialect levelling 2, 5, 32, 43-46, 231-258,344 and morphological inventory 300-303 and paralexification 41 and reanalysis 2, 33, 43-46, 342, 344 as relabelling 39, 44, 60 definition of 2, 25, 38-42 in creole genesis 3, 42, 46-48, 68, 101 in second language acquisition 42, 301 representation of 39-41, 45 resumptives in relative clauses 102-104 in the context of extracted subjects 102 S. semantic transparency of creole languages see creole languages of Creoles' lexicons 183-187 of creoles' morphology 188-196 of creoles' syntax 196-198 of interpretative facts 198-203 Semantic Tranparency Hypothesis 4, 181-182,303-308 evaluation of 204-206 serial verbs see parameters subject raising see verbs see raising T. tense complex tenses 97 markers 95-98 tense, mood and aspect markers 95-98 anteriority 96- 97 complex tenses 97 definite future 96-97 imperfective 96-97

358

ISSUES IN THE STUDY OF PIDGIN AND CREOLE LANGUAGES

indefinite future 96 interpretation of 199-200 inventory of 96, 211 perfective 97 subjunctive 96 tensed complements of control verbs 77-80 of verbs of the SAY-class 98-100 of verbs of the WANT-class 98-100 U. UQAM projects 29-31, xii-xiv V. verb-doubling constructions causal adverbial 201, 213 factive 201-202, 213 temporal adverbial 201, 213 verbs 67-84 BODY-state 71-72 case assigning properties of 83 control 75, 77-80

double object 83-84 existential 70, 77 inherent object 81-83 licensing expletive subjects 70, 74-75 light verbs 81-82 raising 75-77 reflexive 74 semantic properties of 67-69, 183-184 syntactic properties of 69-84 WEATHER verbs 70, 72-74 see argument structures W. Whquestions 146-147 words and phrases 62-67, 146-147 word order 196-198 establishing word order in creole genesis 46, 226 in compounds 51 in nominal structures 88-89

Issues In The Study Of Pidgin And Creole Languages. (Studies in Language Companion)